ISSN 0970-3993
SPECIAL FEATURES:
* Computers in Banking
* Computer Mouse
* Personal Computer Decisions
* Speeding-up the Cpmputer
PROJECTS:
* 3 Vi - Digit SMD Voltmeter
* Intruder Alarm
* DC-DC Power Converter
* Active Loudspeaker Crossover Filters
* Automatic Outdoor Light.
Volume-7, Number-12
December 1989
Publisher : C.R. Chandarana
Editor: Surendra Iyer
Circulation : Advertising : J. Ohas
Production: C.N. Mithagari
Address:
ElEKTOR ELECTRONICS PVT. LTD.
52, C Proctor Road, Bombay -400 007 INDIA
Telex: (011)76661 ELEK IN
OVERSEAS EDITIONS
Elektor Electronics
(Publishing)
Down House, Broomhill Road,
LONDON SW18 4JQ
Editor: len Seymour
Elektor sari
Route Nationals; Le Seau; B.P. 53
59270 Bailleul - France
Editors: D R S Meyer;
G C P Raedersdorf
Elektor Verlag GmbH
Siisterfeld-StraBe 25
5100 Aachen - West Germany
Editor: E J A Krempelsauer
Elektor EPE
Karaiskaki 14
16673 Voula - Athens - Greece
Editor: E Xanthoulis
Elekluur B.V. P"-
Peter Treckpoelstraat 2-4 \
6191 VK Beek - the Netherlands
Editor: PEL Kersemakers
Electro-shop
35 Naseem Plaza
Lasbella Chawk
Karachi 5 - Pakistan
Manager: Zain Ahmed
Ferreira & Bento Lda.
R.D. Estefania, 31-1°
1000 Lisboa-Portugal
Editor: Jeremias Sequeira ^
Ingelek S.A.
Plaza Republics Ecuador
2-28016 Madrid - Spain
Editor: A M Ferrer
Electronic Press AB
Box 5505
14105 Huddinge - Sweden
Editor: Bill Cedrum
The Circuits are domestic use only. The submission ot
designs or articles implies permission to the publisher
to alter and translate the text and design and to use the
contents in other Elektor Publications and activities
The publishers cannot guarantee to return any material
submitted to them. Material must be sent to the Holland
address (given above). All drawings, photographs,
printed circuit boards and articles published in elektoi
publications are copyright and may not be reproduced
or imitated in whole or part without prior written
permission of the publishers.
Patent protection may exist in respect of circuit devices,
components etc. described in this magazine.
The publishers do not accept responsibility for failing to
identify such patent or other protection.
■ty*.
Printed at : Trupti Offset Bombay - 400 01?
Copyright ® 1989 Elektuur B.V.
CONTENTS
Special Features
Super sun-storm management 12.10
Robots for satellite repairs 12.10
Video phones & HDTV 12.11
Remote diagnosis 12.11
Banking Computers 12.15
Audio & Hi-fi
PROJECT: Active loudspeaker crossover filter (2) 12.15
Components
Practical filter-design Part 10 (Final) 12.30
Computers
Computer mouse 12.18
Personal computer decisions 12.46
Speeding up the computer 12.48
Design Ideas
Protecting asynchronous motors 12.36
General Interest
PROJECT: The digital model train Part-8 12.20
PROJECT: Automatic outdoor light 12.32
PROJECT: Intruder alarm 12.33
PROJECT: DC-DC power converter 12.42
a
Radio & Television
Travelling wave tubes 12.60
Science & Technology
Intelligence intentionally & self awareness 12.38
Test & Measurement
PROJECT: 3'A digit SMD voltmeter 12.25
elektor india dccember 1989 12.03
COMPUTER MOUSE
J. Ruffell
Raptly looking at the screen and cheerfully moving the mouse
around on our desks to make our way through menus, few of us
appear to be aware of the operation of the most popular pointing
device for computer applications.
A computer mouse is also called a pointing
device because it allows the cursor (usually
an arrow or crosshairs) to be moved across
the computer screen. You use your hand
to control the direction and speed of the
cursor. Many mouse-oriented programs
allow you to select an option from a menu
on the screen simply by pointing at it and
clicking a button on the mouse. The mouse
has become so popular because it obviates
keyboard commands that distract the at-
tention from the screen and are relatively
slow and susceptible to errors. Another
major application of the computer, draw-
ing, would be unthinkable without a
mouse.
Principle of operation
One aspect common to all computer mice
is that movement is converted into signals
that can be handled by a computer. This is
achieved basically as shown in Fig. 1. An
auxiliary spindle presses a small ball
lightly against two spindles that are
mounted at right angles to each other. Its
own weight, and in some cases the auxil-
iary spindle also, keeps the ball in contact
with the desk surface or mouse pad. The
movement of the ball is hardly obstructed
because the areas where the spindles
touch the ball are small. The friction is,
however, sufficient to cause the spindles
to rotate if the ball is moved horizontally
( x component) or vertically (y component)
in a two-dimensional plane. In this man-
ner, the spindles extract the horizontal
and vertical components from the mouse
movement. These two components are
converted into four electrical signals. This
is done by mounting a slotted disk on to
each spindle. The slots are arranged such
that the light beam of one optocoupler is
fully passed when the other optocoupler
is about half way open. As the spindle
rotates, the optocouplers produce two rec-
tangular signals with a phase difference of
90°. The direction of travel of the spindle
(in one plane) can be deduced from the
phase relation of the two signals. Tire
number of periods of the rectangular sig-
nal indicates the relative distance covered
by the ball, and its speed.
Figure 3 shows how the two rectangu-
lar signals are used to deduce the direc-
tion of travel of the mouse. One
optocoupler signal is called reference, the
other direction. The reference signal deter-
mines the instant the minimum step size
(distance travelled) is reached in the direc-
tion indicated by the direction signal. This
instant is marked by one of the level tran-
sitions (pulse edges) of the reference sig-
nal. Since most computer interrupts are
called by negative pulse edges, it is con-
venient to look at the l-to-0 transition of
the reference signal. As shown in Fig. 3a,
the direction signal is logic high at the
negative edge of the reference signal. For
the opposite direction, however (Fig. 3b),
the direction signal is low at the negative
edge of reference signal. In terms of pro-
gramming, this means that the number
representing the cursor position on the
screen must be changed on the falling
edge of the reference signal. In this soft-
ware routine, the direction signal must be
read to determine whether the cursor po-
sition must be incremented or de-
cremented at a particular step size, e.g.,
one screen position. If, after first connect-
ing a mouse and installing the software
12.18 elektor India december 1989
Fig. 2. Slotted discs and optocouplers are
used to digitise ball movement.
driver, the cursor movement is opposite to
that of the mouse, the reference and direc-
tion signals probably need to be swapped.
The above description of the basic
operation of a mouse applies, at least in
principle, to most other pointing devices
that allow the user to control the cursor
position on the screen direct by moving
the mouse accordingly. There are, how-
ever, also applications that 1 quire a dif-
ferent approach. Take, for instance, a
program that enables a drawing on paper
to be copied into the computer by means
of a mouse. In this case it is the drawing,
not the computer screen, that determines
the cursor position. This type of mouse is
known as a digitiser, and is usually sup-
plied with a special pad. The paper is in-
serted between the digitiser and the pad.
The window in the digitiser 'sees' the pad
surface through the paper. Because the
pad 'communicates' with the digitiser, an
output signal is available that enables the
computer to determine the absolute posi-
tion above the pad, and, of course, above
the paper, which is secured on it. Lifting
the digitiser and putting it down again a
little further is therefore perfectly accept-
able, since the new position is detected
immediately. This is in contrast with a
ball-type mouse, which can not supply
positional information if it is lifted from
the desk.
Another system to convey positional
information to the computer is a combina-
tion of a graticule pad and a mouse with
built-in reflection sensors. The internal
operation is functionally similar to that of
the discs and spindles in the ball-type
mouse. The optocouplers are replaced by
sensors that detect the light reflected by
the pad. The function of the discs is taken
over by the pad with its pattern of light
and dark areas. Like the ball-type mouse,
the optical mouse produces a reference
and a direction signal. Its clear advantage
is, of course, the absence of moving parts.
However, the optical mouse also has its
disadvantages: these are mainly that the
pad has to be kept clean, and that the
pattern on it is critical.
To the computer
The simplest way to convey the rectangu-
lar output signals supplied by the mouse
is, of course, by means of a cable. The
computer has either a built-in mouse
adapter ('bus mouse', e.g. the Amstrad
PC1512/1640 series), or a standard RS232
serial port to which a mouse with built-in
'intelligence' can be connected (e.g., most
standard IBM PCs and compatibles). The
latter mice are often microcontroller-
driven, and supplied with a special soft-
ware program, called the mouse driver,
that enables the PC to translate data re-
ceived at high speed via the RS232 port to
be translated into cursor movement. The
current required for powering the circuit
in the RS232 mouse is obtained from the
computer's serial port. This is possible
only by virtue of the low current drain of
the serial mouse.
The latest in pointing device technol-
ogy is the wireless mouse, which com-
municates with the computer via an
infra-red link. Position output and the
way the data is processing in the driver
are, however, not different from those of
the conventional 'mouse with tail'.
Signal processing
As already stated, the mouse signals are
usually processed by means of a driver
program installed on the computer. Most
computer users will content themselves
with being able to automatically install
the mouse with the correct parameters as
partof the system configuration programs
called at power-on. For advanced applica-
tions, however, mouse manufacturers like
Genius supply a programming guide and
auxiliary programs (e.g.. Genius Menu
Fig. 4. Serial mouse with on-board CMOS microcontroller to guarantee a low current drain
from the RS-232 port on the computer.
Fig. 3. The phase relation between the ref-
erence and direction signals is used to de-
duce the direction of travel.
Maker) that give the user the opportunity
to implement his own pull-down menus
and mouse control in a particular pro-
gram.
Among the many functions of the
driver or the microcontroller in the serial
mouse is adaptive resolution control, or con-
trol of the step size as a function of mouse
speed. If the mouse speed exceeds a cer-
tain predefined value, the cursor step size
is automatically increased. The advantage
of this system is that a relatively small
mouse movement enables large distances
to be covered rapidly on the screen.
elektor india december 1989 1 2.19
THE DIGITAL MODEL TRAIN - PART 8
by T. Wigmore
Construction & testing
IC sockets may be used, but it should be
noted that this is no longer accepted prac-
tice, at least as far as standard logics cir-
cuits are concerned. Some sockets are more
expensive than the IC itself and, more
importantly, the reliability of a circuit is
inversely proportional to the number of
connexion s. None the less, for the more
expensive ICs, such as the A-D converter
(1C25) and the EPROM (IC13), a good-
quality socket is recommended. Bear in
mind also that the printed-circuit board is
through-plated: any desoldering of ICs is,
therefore, a tricky operation. So, check and
double-check whether the 1C is the correct
one before soldering it on to the board.
The parts list shows ICs of the HC- and
HCT-type. The HC-types may be replaced
by HCT-types, but FICT-types should NOT
be replaced by HC-types.
Power supply. Start by fitting D38-D41,
D36, C24, C25 and C27. Next, fit IC29 on to
the relevant heat sink and mount the re-
sulting assembly on to the board. There are
tracks underneath the heat sink that are
protected by a thin layer of lacquer only: it
is therefore necessary to give these extra
insulation (by, for instance, a suitably-sized
piece of thin cardboard or old PCB or insu-
lating tape). The IC should be fixed to the
heat sink with an M-3 bolt, nut and wash-
er, and a generous amount of heat con-
ducting paste.
Connect the mains transformer to the ~
terminals on the PCB. If you intend to use
more than It) keyboards in addition to the
main board, a transformer of higher rating
than indicated in the parts list must be
used, or the keyboards (dealt with in Part
9) must have a separate power supply.
Assuming that the keyboards will be fed
by the present supply, wire link A must be
fitted.
It is possible to use a suitable mains
adapter provided this delivers 9 V at not
less than 800 mA. If the adapter delivers a
direct voltage, D39 and D40 may be
replaced by wire links and D38 and D41
must be omitted.
Switch on the mains and check that the
output voltage of IC29 is 5 V ±5%. If it is
not, disconnect the mains, discharge C25
via a 100 U resistor, and check all the com-
ponents and the preceding work thorough-
ly. If the output is all right, switch off the
mains and discharge C25 via a 100 Cl resis-
tor.
Oscillator. Fit 1C8, IC21, R2, R3, C22, C37,
C40 and the crystals on to the board.
Switch on the mains and verify that a sym-
metrical signal of 2.458 MHz exists on pin
12, and a signal of 614 kHz on pin 8 of IC8.
Microprocessor. Fit 1C4, R8, R12, R18, R19,
R24, C34, D34 (observe polarity!), Tl, IC24,
R13 and C23. These components constitute
the power-up reset for microprocessor IC4.
The operation of IC4 is tested by placing
an instruction on the data bus by means of
hardware. In the first instance, this is the
STOP instruction (76jq: 011 1 011 0g). For
Parts list
Resistors:
Ri = 100Q
R2;R3 - 4k7
R4;Rs;Rii;Ri2;Ri7-R2o;R22;R23;R24= 10k
R6;Rio = SIL resistor array 10k
R7;R8;Ri5 = 3300
R9;Ri4;Ri6 = 47k
Ri3 = 15k
R 21 = 6k8
Capacitors:
C 1 -C 16 = lOn (pitch 5 mm)
Ci7 = 47p
Cib;Ci 9= lOOp; 25V
C 2 o;C 2 i = 220n
C 22 = 33p
C23 = 4p7; 6V3; tantalum
C24;C27 = 470n
C 28 -C 42 = lOOn (pitch 7.5 mm)
C 25 = 2200p; 16V; axial
C 26 = lOg; 6V3; tantalum
Semiconductors:
□i-D32;D37 = 1N4148
D 33 = green LED
D 34 = red LED
D 35 = yellow LED
D36 = 1 N4001
D38-D41 = 1 N5401
Ti;Ta = BC557
T 2 = BC547
ICt = 74HC(T)245
IC 2 = 74HC(T)74
IC3 = Z80PIO (Z8420 or Z84C20)
IC4 = Z80CPU (Z8400 or Z84C00)
IC5;IC6 = 74HCT238
IC7 = 74HCT139
ICs = 74HCT93
IC 9 = MCI 489 or SN75189
IC 10 = MCI 488 or SN75188
ICi 1 ;IC26 = 74HCT32
IC 12 = Z80CTC (Z8430 or Z84C30)
I C 1 3 = 2764 (ESS572)
IC 14 = 6264
IC 15 = 78L12
IC 16 = 79L12
ICi7;ICi9 = 74HCT174
ICis = 4066
IC 20 = 74HCT244
IC 21 = 74HCT04
I C 22 ; IC 23 = 74HCT374
IC24 = 74HCT74
IC 25 = ADC0816
IC27 = MCI 45026
IC 28 = 74HCT1 38
IC 29 = 7805
Note: ICs from the HC-series may be re-
placed by HCT-equivalents. Do not use a
HC type if a HCT type is stated. LS-types
are not suitable because of their higher cur-
rent consumption.
Miscellaneous:
Ki-Kis = 5-way 180” DIN socket for PCB
mounting .
36 off M2x5 screws for securing Ki-Kia.
K 19 = 20-way SIL female header; angled:
0.1 -in. pitch (e.g., Assmann AWRF A20Z).
K 20 = 9-way feamle sub-D connector;
angled; for PCB mounting.
2 off M3x8 screws for securing K 20 .
K 21 = optional 40-way for future extensions.
REi = OIL reed-relay; 5 V coil voltage; e.g.,
Siemens V231 00-V40O5-AO00.
Xi = quartz crystal 4.9152 MHz.
51 ;S3 = push-to-make button.
5 2 = push-to-break button.
Heat-sink for IC 29 : size 30x37.5 mm (e.g.,
SK09 from Dau Components/Fischer).
Mains transformer 8 V or 9 V @ 1 A min.
sec.
PCB Type 87291-5
Additionally required for each loco controller
(max. 16 allowed):
Loco controller:
Potentiometer 100k linear (rotary or slide
type) with knob.
5-way DIN-plug; 180”.
One (EEDTS) or two (Marklin-system) SPST
switches.
Loco address settings (4 options):
1) fixed address setting:
diodes 1N4148, max. 6
2) variable address setting:
8 diodes 1N4148 and 1 8-way DIP
switch block.
3) variable addresss setting:
8 diodes 1N4148
1 6-way header with 2x8 contacts in
0.1 -in. raster,
max. 6 jumpers
4) extra-flexible address setting:
as option 3 but instead of jumpers:
16-way flatcable connector
2 BCD-encoded thumbwheel switches
' number of sockets depends on number of
connected loco controllers. Socket K18 is
preferably a 6-way type for PCB mounting.
12.20 elektor india december 1989
N XXXXZXXX
Fig. 49. Operation of the microprocessor is test-
ed by instructions on the data bus formed by
resistors. The STOP instruction (01110110b) ' s
formed as shown at the top, and the NOP
instruction (00000000) as shown in the lower
illustration.
this, eight 4k7 resistors are connected as
shown in Fig. 49a to where later (possibly)
K21 will be connected. When the mains is
switched on, D34 should light. Switch off
the mains and place the NOP instruction
(00000000) on to the data bus as shown in
Fig. 49b. Switch on the mains and check
the data bus for any short-circuits. Pin AO
should have a symmetrical square wave of
307 kHz; A1 one of 307/2 kHz; A2 one of
307/4 kHz; and so on up to A15, which
should have one of 9.375 kHz.
Fig. 50. Component layout of the double-sided,
through-plated main printed circuit board. The
board is illustrated here on a scale of 95:100.
00900000000 0000000 00'
,00000000000000000000
I xzxzzxzz
>
2
1W»5
SSO&ODBSOSDSBOO&ObOl
aaaaaai
.aaaaaaa. 9
ICll «
Si?
icis
‘oeoooog 6
aaaaaai
im
• mil
aaaaaaaa,
IC28 4
| IC27
J
qgpoooc
r
«ror
elektor india december 1989 1 2.21
Memory. The next step is the mounting of
the EPROM (IC13) that con tains the con-
trol program, the RAM (104) and the
memory address decoder (IC28). At the
same time, fit decoupling capacitors C33,
C35 and C36. Next, fit IC3, 102, Rll, R16,
R17, R9 (immediately adjacent to C25),
R22, R15, D35, T3, IC7, IC26, C32, C41,
C42, SI and S2.
Switch on the mains and press SI,
when the program should go into the ser-
vice routine, indicated by the flashing in a
1 Hz rhythm of D35. If this happens, 1C3,
102, IC4 and the memories work satisfac-
torily. If, however, D33 lights, the control
program has gone into the internal RAM
test routine: this is almost certainly caused
by 103 and associated components.
Serial output. Fit IC11, 107, 108, IC23,
IC27, C30, R7, R14, D33 and T2. Switch on
the mains and press SI: a low-frequency
square wave should then be present at
pins QO to Q7 of IC23. The frequency of
that signal at QO should be 1 Hz and that
at successive output pins should be one
half of that at the preceding pin.
Pin QO becomes alternatively high and
low every half second; Q1 every second;
Q2 every two seconds; and so on. These
frequencies were chosen this low to enable
them to be checked with an ordinary mul-
timeter. A similar check must be carried
out at the outputs of IC17. Again, the first
output becomes alternatively high and low
every half second and the last one, Q6,
every 16 seconds. Note that D35 flashes in
unison with output QO of IC23, and D33 in
unison with Q6 of IC17.
±12 V supply. The +12 V supply is used
not only for the RS232 interface, but also
for the booster. It is, therefore, required
even if the RS232 interface is not used.
Fit C18-C21, IC15 and IC16. The input
voltage for the supply (±20 V) is taken
from the booster board (see Part 6 -
September 1989) and connected via K17.
This connector is shown in the parts list as
a 5-way DIN socket, but a (hard-to-obtain)
6-pin type is preferred, because this pre-
vents the connecting cable from being
plugged into one of the other DIN connec-
tors by accident. Because of the presence of
the ±20 V potentials that would almost cer-
tainly have disastrous consequences.
The wires in the cable between the
main board and the booster board must be
connected to identically-numbered pins on
K1 and K17. If a 6-way type (which has
different pin numbers) is used for K17,
stick to the numbers given on the boards.
Switch on the mains to the booster unit
(NOT to the main board). The potential at
pin 1 of K18 (with respect to pin 2) should
be -20 V and that at pin 3 (again with
respect to pin 2) should be +18 V. The out-
put voltage of IC15 should be +12 V and
that of IC16, -12 V.
A-D converter and locomotive address
decoder. Fit Rl, R4, R5, C26, C31, C38, IC1,
IC2, IC25 and resistor-array R6. Instead of
an array, eight 10 kfi resistors may be fit-
ted vertically as shown in Fig. 51. Note
that the common earth connexion must be
at the underside.
eight 10 kU resistors may be fitted vertically.
To enable writing the loco addresses
associated with the loco controllers, IC6
and (if more than eight loco controls will
be used) IC5 are needed. Loco controls
may then be connected to K9-K16. The
controller with the highest connector num-
ber has the highest priority if the addresses
are coded identically. In other words, if in
positions 10 and 14 the controllers have the
address 00, that in position 14 will have
priority over that in 10.
Construction of a loco controller. The A-D
converter can not be tested until a loco
controller is available. From a circuit point
of view, these controllers are fairly simple:
three possible designs are shown in Fig. 52.
For each of these designs a 5-way DIN
plug (180°), a 100 kQ potentiometer and
one or two switches are required. Note that
the housing of the DIN plug is used as the
sixth (earth) pin.
It is possible to connect the loco con-
trollers direct to the main board, i.e., with-
out plugs and sockets. This is a particular-
ly logical (and less expensive) method for
controllers that are to be built in perma-
nently.
Each loco controller is associated with
one or two switches for the switching on
and off of the controller, the setting of the
type of data format and, possibly, the addi-
tional decoder switching function.
If a mixture of Elektor Electronics and
Marklin loco decoders is used, the con-
troller design shown in Fig. 52a should be
used. The design in Fig. 52b is intended for
Elektor Electronics controllers and that in
Fig. 52c for Marklin or the modified Elek-
tor Electronics controller (see Part 3 - April
1989).
A controller is considered to be out of
action if both pin 4 and pin 5 of the DIN
connector are open and therefore also if the
relevant DIN connector on the main board
is not connected up.
Switch SI in Fig. 7b and 7c may be
replaced by a wire link at the relevant DIN
connector. A controller can then be taken
out of action only by removing the plug
from the DIN socket.
If the connexions between the main
board and the controllers are fairly long, it
is recommended to use screened cable.
Each loco controller needs a filter capa-
citor and two diodes, all of which may be
fitted on the main board.
Diodes D1-D32 must be fitted vertical-
’I f 1
87291 -VII -17
H I
87291 -VII -18
Fig. 52. Three possible designs of a locomotive
controller. Choice of the design depends on the
type of locomotive decoder used.
12.22 elektor india december 1989
Setting the loco addresses. In general, loco
addresses must be presented in BCD for-
mat as shown in Fig. 54. Valid addresses
are in the range 00—80 (note that Marklin
does not count 00 as a valid address).
Invalid addresses are simply ignored. A
number of possibilities of setting the ad-
dresses is shown in Fig. 56.
The method of Fig. 56a is by far the
least expensive, but has the disadvantage
that addresses can be changed only with
the aid of a soldering iron.
The method in Fig. 56b is the one used
in the present design. The DIL switches
permit setting and altering the addresses
at any given moment, even during opera-
tion of the system.
It is also possible to program the loco
addresses via the RS232 port: this method
will be discussed in a later instalment.
87291 -VII -19
Fig. 53. Possible design of a front panel
for the loco controllers.
□
□
□
PF
□
□
□
x 80
40
20
10 /
8
4
2
1 y
\
7 Y
example of
locomotive
address =
= 59 (40+10+8 + 1)
4 «:
connexion v
open
Fig. 54. Loco addresses (00-80) must be presented in BCD
format.
J 16-way header
shorten _
flatcable
digit 2
tens
digil 1
units
□
□
□
□
□
□
> digit 2
_ — digits 1 & 2 common
Fig. 55. Thumb-wheel switches may be connected via flatcable.
Unused wires should not be connected to prevent unnecessary
capacitive loads.
ly.
Since the DIN sockets are subject
to fairly large mechanical strains dur-
ing the insertion and withdrawal of
plugs, they should be fixed to the
board with M2x5 nuts and bolts or
with small self-tapping screws before
the solder connexions are made.
Loco controllers and the A-D con-
verter may be tested by connecting
them to K16, which is the most impor-
tant loco controller socket. The setting
of the loco addresses will come later:
for the time being, they will be written
as 00.
Switch on the mains to the main
board, but do NOT press SI. The nor-
mal control program will then be
active. After a moment or two press SI
when D33 should light. Also, the sig-
nals resulting from the A-D conver-
sion are present at outputs D3-D7 of
IC25, while at pins 6 and 9 of IC2 the
switch position may be verified: if the
output is 0, the switch is closed and if
it is 1, the swatch is closed.
Output relay. Fit Rel, D37, IC10, R20,
R21 and C29. When the mains is
switched on, pin 3 of IC10 should have
a d.c. potential of -10 V to -12 V. When
SI ('go') is pressed, the output relay
will be energized in unison with the
lighting of D33.
Also, the same potential as at pin 3
of IC10 should be present at pin 4 of
K17. When in this condition a loco
controller is connected, the potential
should vary slightly when the poten-
tiometer is adjusted. The degree of the
variation depends on the loco address.
This voltage is no longer a true d.c. poten-
tial as may be verified with an oscillo-
scope, which will show the repeatedly sent
loco control instructions whose rear por-
tion varies according to the position of the
potentiometers and function switches,
while their front portion varies according
to the relevant loco address.
Fig. 56. Four possibilities of setting loco addresses: (a) with diodes (address = 48); (b) with diodes
and DIL switches (address = 21); (c) with diodes and shorting plugs (address = 42); (d) with diodes Keyboard interface. This section of the
and thumb-wheel switches (address = 71). board need ' of course ' onl y be P°P ulated > f
clektor india december 1989 12.23
it is intended to connect keyboards (which
will be dealt with in next month's instal-
ment) to the main board.
Fit resistor-array RIO (but see Fig. 51),
R23, C28, IC19, IC20, IC21 and K19. The
choice of a single-in-line type for K19 was
deliberate, because if the keyboards are
installed permanently, they may be con-
nected by means of wire links instead of
by relatively expensive plugs.
RS232 interface. To populate the last sec-
tion of the main board, fit IC9, Cl 7, K20
and K18.
The installation of the main board is left to
your own requirements, but bear in mind
that keyboards must be connected to the
left-hand side of the (flat) case
Some operational tips
Loco controllers are scanned from left to
right. If several controllers are set to the
same address, the one at the extreme right
will have priority over the others.
As in the Miirklin system, it is possible
to set the speed of one locomotive with a
given controller and then use that con-
troller for a different loco address, without
affecting the operation of the first loco.
If the mains is not connected to the sys-
tem and SI is pressed, the green LED
(D33) will light, but go out as soon as SI is
released.
The system can not and will not send
data until the booster is switched on and
the go key (SI) has been pressed. If the
connexion with the booster is broken, the
system will automatically come to a halt.
The system ignores brief (< 0.5 s) short-
circuits. Again, if the system switches itself
off, it may be reactuated by pressing SI.
In emergencies, the system may be
stopped by pressing S2: this not only inca-
pacitates the control program, but it also
removes the power from the rails. If
desired, a number of these stop switches
may be installed in series along the track.
Switch S3 is the system reset control,
which normally will not be used. Only if
the system does not appear to react to any
other control or if D34 unexpectedly
lights, should this switch be used.
NEW PRODUCTS
TV Picture Cleaner
Say goodbye to disturbances in your TV
Picture caused by tubelights, grinders,
belnders, autorickshaws, scooters and
low-flying aircraft. The TV Picture
Cleaner, when connected to antenna
cable, removes them electronically. The
product can be easily connected to the
antenna cable. Used throughout the
world, this device has been designed and
manufactured in India by Magnum Elec-
tric.
Magnum Electric Company Pvt. Limited
• 2, Ramavaram Road • Manapakkam •
Madras- 600 089. Phone: 434547.
Microprocessor Controlled
Charge Amplifier
voltage signal. Main characteristics are
continuous range setting from ±
10. . .999,000 pc and LCD parameter set-
tings. The parameters remai set if mains
failure occurs. Frequency range from 0-
200 kHz, automatic zero offset correc-
tion, in built low pass filter with 8 select-
able cut off frequencies, 3 selectable
time constants. An IEEE-488 Interface
enables the Charge Amplifier to operate
via a Computer. Available against Ac-
tual Users import Licence or Open Gen-
eral Licence as applicable.
M/s. Integrated Process Systems • 9,
M.P. Avenue • Santhome • Madras- 600
004.
features an In Circuit Low Voltage Tes-
ter. Other attributes include a Con-
tinuity Tester, Data hold facility, Diode
Tester and add-Ons to measure high vol-
tage upto 30 KV DC/20 KV AC. (HV
Probe), Currents upto 20 A (Current
Shunt), Frequencies upto 20 KHz (A.F.
Probe) and 10 MHz.
Pla Electro Appliances Pvt. Ltd. •
Thakor Estate • Kurla Kiri Road • Vid-
yavihar (West) • Bombay- 400 086.
The mains operated, microprocessor
controlled single channel; charge
Amplifier Type 5011 from our Principals
M/s. KISTLER INSTRUMENTS AG,
SWITZERLAND, is a new concept in
converting the charge yielded by Piezo
Electric Transducers into a proportional
Digital Multimeter
Pla has introduced a 4V4 digit LCD digi-
tal multimeter. This accurately measures
AC/DC voltage, current and a broad
range of resistance. The instrument also
12.24 elektor india december 1989
31/2-DIGIT smd voltmeter
T. Wigmore
This little circuit is simple to build, offers good accuracy and can be
used in all applications requiring a small voltmeter with a clear LED
read-out.
Much of today's electronic equipment re-
quires a digital read-out to show system
status or process variables. Such read-
outs are usually compact voltmeter mo-
dules with an LC (liquid crystal) display.
The present read-out is also a voltmeter,
but uses displays with light- emitting
diode (LED) segments. A LED indication
was chosen for this application because it
remains visible in the dark (this require-
ment would also have been met by an
LCD with back-lighting). Also, the use of
7-segment LED displays in combination
with a drive circuit built with SMA (sur-
face-mount assembly) components allows
a really compact voltmeter to be realized
— see Fig. 1. This is particularly import-
ant if the meter is to be built into existing
equipment.
One integrated circuit
The circuit (Fig. 2) is formed by a single
integrated circuit Type ICL7107 from In-
tersil. This voltmeter IC is the LED version
of the perhaps even more familiar
ICL7106 for LCDs. The ICL7107 contains
everything required for the analogue-to-
digital conversion of the input signal, and
the driving of a 3VS-digit read-out. The
chip is used in a more or less standard
application circuit with some extra com-
ponents to afford flexibility as regards the
power supply.
Analogue-to-digital
conversion
Analogue-to-digital (A-D) conversion can
be accomplished in a number of ways.
Fast converters almost invariably use
Fig. 1. The compact voltmeter module seen at different viewing angles.
elektor india december 1989 12.25
314>-DIGIT SMD VOLTMETER
•
Read-out:
3Vfe-digit LED display
•
Sensitivy:
+200 mV; differential input with symmetrical supply
•
Decimal point:
2 positions; indication 188.8 or 18.88
•
Reference:
internal or external
•
Supply voltage:
single 5 V (limited common-mode);
5 V with negative bias;
symmetrical (±5 V)
•
Current consumption:
max. 200 mA from positive (+5 V) supply;
300 pA from negative supply
•
Size:
55x37x11 mm
flash ADC chips that are characterized by
a large number of internal comparators.
The other principle, successive approxi-
mation, is based on a resistor ladder net-
work whose R-2R junctions are connected
to counter outputs. The result of the D-A
conversion is compared to the input sig-
nal. If a difference is detected, the clock
oscillator with the counter is controlled
accordingly until the output voltage of the
internal D-A converter equals the exter-
nally applied voltage. In practice, the ac-
curacy of this type of converter is that of
the R-2R network, and the off-set voltage
of the voltage comparator.
The ICI-71 07 and other ICs in its family
work on yet another principle, w’hich is
entirely analogue and based on an inte-
grator. Internal off-set voltages are com-
pensated prior to any measurement cycle,
so that a high accuracy is achieved even
with small input voltages. Since the meas-
urement principle is based on the com-
parison of an input voltage, U i, with a
reference voltage, Uref, the display value
is in fact Ui/Umi. Interestingly, the refer-
ence voltage may be applied externally.
Three phases
The measurement cycle of the 1CL7107
consists of 3 phases. Figure 3 shows the
signal path in the analogue input circuit
for each of these.
During the auto-zero phase (Fig. 3a),
inputs IN LO and IN HI are disconnected.
Internally, a closed loop is formed consist-
ing of input buffer amplifier A i , integrator
A 2 and comparator Ar (Gnt is discharged
as yet). The internal ground is formed by
the analogue common potential. The auto-
zero capacitor will charge to a voltage that
compensates the off-set voltages of Ai, A’
and A 3 . Also, Cref is charged to the refer-
ence potential.
The auto-zero phase is followed by the
integration phase. The input voltage be-
tween IN LO and IN HI is applied to an
integrator formed by Az-Rmt-Cint. The in-
tegration interval is defined as 1,000 clock
cycles. During this interval, the output
voltage of the integrator rises to a value
directly proportional to the input voltage.
The last phase is the de-integration
phase. The input voltage to the integrator
is disconnected again and replaced by the
voltage on Crd. An internal circuit allows
the reference voltage to be connected with
the opposite polarity of the previously ap-
plied input voltage. This causes the inte-
gration process to be reversed, and the
interval to be timed by the internal clock.
The number of clock pulses is directly pro-
portional to the ratio of the reference volt-
age to the input voltage. This principle is
best understood by assuming the refer-
ence voltage to be equal to the input volt-
age, which results in a de-integration
phase that is just as long as the integration
phase. The length is 1,000 clock cycles,
which is shown on the display. If the input
voltage is only half the reference voltage,
the de-integration process takes half the
time of the integration process, and the
display will read 500 to indicate that Uin =
O.SOOUref.
The length of the de-integration phase
depends on the input voltage. With rela-
tively long de-integration phases, the
auto-zero phase is automatically short-
ened so that the total measurement time
— and with it the number of read-outs per
second — remains constant. The integra-
tion phase always lasts 1,000 clock cycles,
the de-integration phase 0 to 2,000 clock
cycles, and the auto-zero phase 1,000 to
3,000 clock cycles. One complete measure-
ment cycle takes 4,000 clock cycles, bear-
ing in mind that the clock frequency is
divided internally by 4. A clock frequency
of 48 kHz gives an internal clock fre-
Fig. 2. Circuit diagram of the voltmeter.
quency of 12 kHz to allow 3 measure-
ments per second.
Common mode
The dual slope measuring principle used
by the ICI-71 07 has been discussed in
some detail to show' up the limitations of
the common-mode arrangement.
Clearly, satisfactory measurements
can be made only if the reference and
input voltages lie within common mode
range, V-(+l V) to V+C-0.5 V), of the in-
ternal amplifiers. Another requirement is
for the integrator output voltage to re-
main well below' the positive supply volt-
age. During the integration phase, the
LD1-.LD3= HD1105
LD4 = HD1108
All components (except displays) SMD
12.26 elektor india december 1989
integration phase 890117- 13
(1000 cycles)
(0 - 2000 cycles)
Fig. 3. Signal paths illustrating the basic three-phase operation of the analogue input
stages of the ICL7107 voltmeter chip ( courtesy GE-Intersil ).
voltages at in lo and in hi are connected
to the inputs of the internal buffer ampli-
fier and the integrator, and must, there-
fore, fall within the common-mode range.
The reference voltage is never applied di-
rect, but via the previously charged capa-
citor Gel. This means that the
common-mode voltage range (CMVR) of
the reference voltage is the supply volt-
age, i.e., V+ to V— .
During the integration phase, the inte-
grator uses the potential at in lo as the
reference. De-integration, however, is ef-
fected with respect to the 'common'
potential. Consequently, any difference
between the in lo potential and the com-
mon potential causes a voltage jump at the
integrator output during the switch-over
from integration to de-integration (see
Fig. 3b).
Displays
In the circuit diagram in Fig. 2, the oscil-
lator frequency is set to 48 kHz by compo-
nents Ci-Ri. This frequency results in
3 read-outs per second, and may be
adapted to individual requirements by
changing Ri-Ci as appropriate, bearing in
mind that the integrator time-constant,
R 2 -C 4 , must be changed at the same time.
Input filter Ra-Cs ensures a stable read-
out.
The segment current capability of 5 to
8 mA of the ICL7107 obviates additional
driver transistors and current limiting re-
sistors. The read-out is composed of
3 common-anode 7-segment LED dis-
plays Type HD1105, and 1 common-ca-
thode display Type HD1108. The latter is
used because l^-digit, 12.7 mm-high, LED
displays are difficult to obtain in com-
mon-anode versions. Fortunately, the ca-
thode of the minus sign on the HD1108 is
not connected to the A and B segment.
Both the HD1105 and HD1108 are manu-
factured by Siemens.
Internal and external
reference
The internal reference source of the
1CL7106 and the ICL7107 may be used
with a sufficiently high supply voltage
(more than 6.5 V between V- and V+). The
temperature characteristics of this refer-
ence may, however, cause problems with
the SMA 1CL7107 because this is a rela-
tively small chip, and drives LEDs direct.
For this reason an external reference, e.g.,
the ICL8069, may be used. Other reference
devices may be used provided R.i is modi-
fied accordingly to ensure optimum bias
current (note that the voltage difference
between ref LO and V+ is typically 2.8 V).
Resistor R- has a value that allows multi-
turn preset Pi to be adjusted to give a
reference voltage of 100 mV between REF
LO and REF hi.
Construction
The printed-circuit board (Fig. 5) accom-
modates the voltmeter circuit and the dis-
plays. The board is cut in two to enable the
display section to be jn minted either ver-
tically or horizontally on to the voltmeter
board.
All components, except the optional
reference, IC 2 , multiturn preset Pi and the
4 displays, are surface mount assembly
(SMA) types.
The values of R 3 and R- depend on
whether or not IC 2 is used, while compo-
nents Rj, C 7 and Di may be required only
with certain power supplies as discussed
below.
The two jumpers on the board allow
the decimal point to be positioned either
between the first and second digit (e.g.,
100.0) or between the second and third
digit (e.g., 10.00). The third option, 1.000,
is not possible because the fourth digit is
a common-cathode type.
Power supply
In most cases, the voltmeter will be incor-
porated into an existing piece of equip-
elektor india december 1989 1 2.27
Fig. 4. Signal waveforms with terminals lo an common connected (top drawing) and with
a potential difference between lo and common (lower drawing) ( courtesy GE-Intersil).
Parts list
C5;C6;C7 = 47n
All parts surface-mount assembly except
Semiconductors:
when marked ♦.
Di = zener diode 4V7; 400 mW
LDi ;LD2;LD3 = HD1 1 05R (Siemens) *
Resistors:
LD4 = HD1 108R (Siemens)*
Ri = 100k
ICi = ICL7107 (GE-Intersil)
R 2 = 47k
R 3 = 4k7
IC 2 = ICL8069 (GE-Intersil) *
R4 = 470n
Miscellaneous:
Rs = 680fi
PCB Type 8901 17
Re = 1 M0
3
II
o>
£
Pi = 1 k0 multiturn preset *
Capacitors:
Cl = loop
C2=100n
C3 = 470n
C4 = 220n
r
o o
Ol
<00000000000 00000 000000000
( 1)0
am
UJ 00
LD4
LD3
LOS
LDI
1
! | — 1
I “|
• /
/ l
/ /
/ /
Cl
, 0.1
Cl.
, ; JP3 ; JP2 ,
boooooooooo 00000 oooooooooj
Fig. 5. Track layout and component
mounting plan of the printed-circuit board.
1 2.28 elektor india december 1389
o
890117-18
o
Fig. 6. Power supply configurations.
ment with an internal power supply.
Without displays, the voltmeter draws
1.5 mA at 6 V max. between V+ and
ground, and -300 pA at 9 V max. between
V-and ground. With displays, the current
drawn from the positive supply lies be-
tween 70 mA and 200 mA, depending on
the number of actuated display segments.
The negative supply need not source more
than 300 pA, and is not even required in
some applications.
The positive supply voltage is limited
to prevent the maximum dissipation of
the ICL7107 being exceeded.
Figure 6 shows the various supply op-
tions. The first drawing. Fig. 6a, shows the
most universal solution based on a sym-
metrical power supply. A 0 fl or other
low-value resistor is fitted in position R-t
(0 Q resistors are quite common in sur-
face-mount technology), and Di is not
fitted.
The circuit of Fig. 6b may be used if a
sufficiently high, regulated, supply volt-
age is available in the equipment. It
should be noted that the input voltage is
not measured with respect to ground.
Another possibility is shown in Fig. 6c.
A single-rail power supply with an output
voltage of 12 V or more may be used if the
negative supply to ICi is limited by fitting
Di and Ra.
In many cases, a single 5 V supply may
be used as shown in Fig. 5d. This applica-
tion requires the use of the external refer-
ence and the fitting of JPi.
Input voltage and sensitivity
In deciding the range of the input voltage,
due account should be taken of the com-
mon-mode voltage. Fit jumper JPi if the
input voltage floats with respect to the
display unit.
Non-floating input voltages must lie in
the range V-(+l V) to V+(-0.5 V). When
the input voltage is close to V-, the read-
out, on going negative, may change sud-
denly to a large value, e.g., -005 instead of
000, -001 etc. This effect may be prevented
by shifting the common-mode input volt-
age towards the middle of the supply volt-
age.
Set the sensitivity to 200 mV full-scale
indication by adjusting Pi for 100 mV be-
tween ref LO and REF HI (the reference
voltage is half the full-scale indication).
The preset allows small adjustments to be
made as required for other sensitivities. If
the meter is to be made less sensitive,
either an external voltage divider must be
fitted, or Pi must be made larger. The lat-
ter solution, however, requires the inte-
grator resistor to be increased accordingly
to prevent clipping of the integrator.
elektor india deccmber 1989 1 2.29
PRACTICAL FILTER DESIGN - PART 10
by H. Baggott
This final part of the series discusses all-pass filters. Strictly speaking,
these networks are not filters since (ideally) they have zero attenuation
at all frequencies. However, they introduce a specific phase shift
or time delay that is very useful in many applications.
Although all-pass networks have zero
attenuation at all frequencies, they intro-
duce a certain phase shift and act, there-
fore, as a sort of delay line. They may be
used, for instance, to delay a signal in
time or to modify the phase behaviour of
an other filter.
A look at the complex field of these fil-
ters shows that their zeros of network
function are mirror images of their poles.
Since the poles are always located to the
left of the v-axis (because of the required
stability of the filter), the zeros must
always be to the right of the ordinate.
Thus, a first-order network is always a real
pole-zero combination.
It is interesting to note that owing to
the unique character of an all-pass net-
work the introduced phase shift is always
twice the value of that of a conventional
filter. The maximum phase shift in a tradi-
tional first-order filter is 90°, while that in
a first-order all-pass network is 180°.
First-order network
The transfer function of a first-order all-
pass network is
T(jw) =
j co - a
j( 0 + a
where a indicates the location of the pole.
The absolute value is
I T (j n>)l=
■yj co' + a 2
a/ of + a
It is seen that for every frequency the
nominator and denominator have the same
value. The associated phase shift is
i p= - 2 arctan ( col a)
The time delay. /, is also important in
all-pass filters; it is calculated from
dy _ 2 a
d w ru 3 + or
The time delay in a first-order network
is always maximal at very low frequencies
and decreases gradually with increasing
frequencies. The gradient of the increase
depends on the value of a. When a is
small, the time delay is large at 0 Hz, but
decreases very rapidly with rising frequen-
cies. When a is large, the time delay is rel-
atively small at 0 Hz, but remains fairly
constant over a wide range of frequencies.
Second-order network
A second-order filter affords rather more
freedom in design, so that the time delay
curve can be matched more accurately to
the requirement.
The transfer function of this type of
network is
2 (O r 2
(j <o) -j co-Q- + w r
T (j ®) = a 7
(jft» +)(0—j- + co'
The absolute value of this function is
again 1 . The presence of the resonant fre-
quency a) r is explained by the fact that
this function concerns a resonant circuit.
This frequency may be calculated from
/ 1
) = -\ cr
r V
+ P
in which a and /J are the poles of the func-
tion.
The Q factor is
Q- (0,12a.
The phase shift of a second-order filter
cp= -2arctan -
coco ,
Q(co r - ft)')
while the time delay is calculated from
2o) 2 ( ft) 2 + ft) 2 )
I =
Q( co' -co') +
(O' (O'
From these formulas, it is clear that the
computations of a second-order network
are not all that simple. The time delay is
largest at the resonant frequency. The
higher the Q. the more pronounced the
peak in the time delay characteristic.
Practical passive networks
The design of a first-order delay network
is fairly simple. Fig. 52 shows two possi-
bilities: a ladder type and an asymmetric
type. Both filters have identical output
type; (b) asymmetric type.
Fig. 53. Time delays of a first-order network at
a-values of 0.1, 1.0 and 10 respectively. Fig. 2a
shows the phase shift and 2b the time delay.
1 2.30 elektor india deccmber 1989
LI
'/5C1
Fig. 54. Circuit diagrams of (a) a second-order
ladder network; (b) an unbalanced network with
a Q > 1; and (c) an unbalanced network with a
Q< 1
impedances, so that they may be cascaded
without any problems. The compuation of
such a filter is quite easy:
L = R/a
C= MaR
where R is the desired output impedance.
The construction of the ladder network
should not present any difficulties, but in
building an asymmetric type it should be
borne in mind that the inductor is centre-
tapped: the magnetic coupling factor
between the two halves must be 1.
The phase shift and time delay curves
given in Fig. 53 are given for a-values of
0. 1. 1.0 and 10. Note that the value of a
may be chosen freely, dependent,; of
course, on the desired time delay curve.
Second-order networks are a little more
complicated and may be designed for Q-
values smaller and greater titan 1. Several
designs are shown in Fig. 54: in (a) a lad-
der network; in (b) an unbalanced filter for
(2-values greater than 1 and in (c) an
unbalanced filter for (2-values smaller than
1 . The designs in (a) and (b) use standard
components throughout, whereas that in
Fig. 55. Designs of active first-order networks: 55a shows a lagging network and 55b a leading one.
Fig. 56. An active second-order network; this design is suitable for (7-values from 0 to 20.
(c) requires a centre-tapped inductor. The
values of the various components are cal-
culated as follows.
The components in these circuits are
calculated as follows.
a 2 + p
C = — 5 —
i 2 aR
V = 2 *> C ,
L =-*-
2 2 a
R (a 2 + ft")
i _ 2L
S ~ a + ~ ; :
or + p
R(p~-3cf)
Active networks
There are even better possibilities of de-
signing active all-pass networks than pas-
sive ones, but for clarity's sake they will
be restricted to first- and second-order net-
works.
Good designs of a first-order filter are
shown in Fig. 55: (a) is an inductive type
and (b) a capacitive type. Furthermore,
both circuits invert the input signal (which
has nothing to do with the phase shift).
Note that not a few people mix up the two
circuits under the impression that the one
in (b) is a lagging type.
(coR jCj) +1
(p= -2arctan( coR p
The design of an active second-order
network is shown in Fig. 56. It consists of
a band-pass filter and a summing amplifi-
er. The computation of the components is
rather more complicated than with the
first-order filter. First we assign a value to
C and then:
R^R } /2
IQ' -1
*3 aC
Next, R5 is given a suitable value, say,
22 k£2. For unity gain, R(> = Rs, but if
amplification is required, Rt> should be
given a larger value.
For 2-values greater than 0.7, Rl is not
required, while
R , = R } /4Q~
elektor india december 1989 12.31
and
With the aid of second-order all-pass
networks, it is possible to design delay
lines that have a constant time delay over
a given range of frequencies. The pole
positions may be obtained from the tables
given earlier in the series. The calcula-
tions are fairly complicated and will not
be gone into here.
Although it is possible to design delay
lines in this manner, the normally specific
requirements of these devices make it dif-
ficult to to give general examples. The
formulas given in this final part must,
therefore, suffice.
automatic outdoor light
shine a light on your door j Bodewes
The purpose of this circuit is to
automatically switch on an outside
light to illuminate your front door,
when a visitor arrives.
The circuit uses a light detecting
resistor (LDR) as the sensor. For the
circuit to work an external light
source such as a lamp post is required.
Needless to say this source needs to
be close by. Please remember that the
removal or repositioning of lamp posts
needs the authority of the local coun-
cil, so we do not recommend this
circuit to anyone who has to
extensively remodel the landscape.
The LDR is mounted into a tube,
behind a lens, and aimed at the light
source. This structure is positioned, so
that the person approaching the front
door, causes a shadow to fall onto the
lens. Do not forget to ensure that the
tube containing the LDR is water
tight. Immediately the LDR is in
shadow, its resistance will increase.
This results in T1 applying a negative
pulse to T2 via Cl and R6. T2 con-
•v*
V
tinues to conduct until this negative
pulse arrives. As soon as T2 cuts-off,
C2 starts to charge. When the voltage
across C2 rises above 2 V, the
schmitt-trigger formed by T3, T4, T5
(and their surrounding components),
switches on transistor T6. T6 conducts
and triggers the relay, which switches
on the outside light. The rate at which
C2 discharges is adjusted by PI . When
the voltage across C2 falls below
1 .5 V the schmitt-trigger returns to a
quiescent state. T6 will cut-off
switching off the relay and therefore
the light.
The light will remain on for a maxi-
mum of one minute. Longer periods
are possible, but then C2 will have to
be substituted with a larger capacitor.
Switch SI and R3 are connectecfin
parallel to R2. SI can be a make/break
contact mounted on the front door.
When the door is opened the light will
switch on, going out immediately it
is shut.
In order for the circuit to work effec-
tively, the tube containing the LDR
(and lens), must be positioned, relative
to the light source, so .that the voltage
measured at the junction of R1 , R2,
is not less than 3 V, and not more
than 20 V.
12.32 elektorindia decamber 1989
INTRUDER ALARM
In today’s society, it makes good sense to provide some form of
intruder alarm system in the home, if for no other reason than the
family’s peace of mind. Effective, reliable and simple to control, the
intruder alarm system described in this month’s article uses readily
available low-cost components only.
E. Chicken, MBE, BSc, MSc, CEng, FIEE
Apart from its low current demand from
a battery during non-alarm conditions,
the alarm is also noteworthy for its sys-
tem-test bleep on switching on and on
leaving the house, its pulse drive of the
external sounder to economize on battery
power, and automatic time-out of the in-
ternal and external sounder to minimize
social disturbance.
The block diagram given in Fig. 1
shows the various stages of the circuit,
their interconnections and related signal
routes. The way in which the stages inter-
act in detail is explained below.
Circuit description
Power supply
As shown in the circuit diagram of Fig. 2,
the alarm is powered by a small 12 V re-
chargeable battery that is trickle-charged
by a mains adapter with d.c. output. In the
quiescent condition, the current drain
from the battery is less than 1 mA. Current
consumption in the actuated condition is
virtually that of the external sounders
alone. Charging current for the rechar-
geable battery is limited to about 15 mA
by R 7 in series with LED D4, which,
mounted on to the front-panel of the en-
closure, serves as a charging indicator.
The output voltage of the mains adapter
must be measured and the value of R7
chosen such that the maximum LED cur-
rent of about 20 mA is not exceeded.
On/off control
Control of the alarm system is effected by
a single-pole ON/OFF switch, Si. Actually,
the circuit is never switched off complete-
ly as long as the battery is connected, but
the current drawn with the switch in the
OFF position is negligible.
Closing Si to switch the system off con-
nects R: to the negative supply rail, caus-
ing Ti to conduct. Diode Di is
forward-biased, and the resultant voltage
drop of about 0.6 V maintains conduction
of T i in the event of a reduction of the
supply voltage. That conduction in turn
maintains the off condition of the system,
and so minimizes the possibility of false
alarms.
When Ti is switched on, Dj ceases to
conduct so thatCi is charged to the supply
voltage via Rs and Ri.. For convenience,
low voltages from 0 to, say, +2 V will be
referred to as logic 0, and the higher +12 V
supply rail voltage as logic 1 .
This voltage on Ci forms a logic 1 that
is inverted by NAND gate Ni to present a
0 to one of the two control inputs of the
bistable formed by N 2 and N.i. So long as
pin 6 of N 2 remains at 0, the output of the
bistable, pin 4 of N 2 , is held at 1 to prevent
the alarm sounders being actuated.
Switching the system off simulta-
neously takes the RESET pins of timers IC2
and IC3 low, which prevents the timers
being inadvertently triggered into a false
alarm sequence. As long as the system is
switched off, D 2 is forward-biased via Ti
and R14.
When the system is switched on,
switch Si is in fact opened, so that Ti
ceases to conduct. This causes the collec-
tor voltage to drop to practically 0 V via
R4, so that D 3 is forward-biased via Rs and
Ri. As a result, Cs discharges slowly via
Rt, D 3 and R4. The lowest voltage on C3 is
reached in about 15 seconds, determined
by time constant CifRi+Rb).
The final voltage on Ci as determined
Fig. 1 . Block schematic diagram showing the general structure of the intruder alarm.
elektor india december 1989 1 2.33
by potential divider Rs-Rr is about one
tenth of the supply voltage, plus the for-
ward drop of D 4 . In total, this makes about
+1.8 V, which represents a logic 0. The
resultant logic 1 at the output of NAND
gate Ni causes bistable N 2 -N 3 to toggle
15 seconds after switching the system on.
The logic state at output pin 4 of the bi-
stable becomes 1, and can be changed to 0
according to the logic level applied to the
control input terminal, pin 1 of N 3 .
Alarm sensing
When all doors and windows protected by
the detector loop are closed, and assum-
ing that the detector switches are of the
normally-closed type, R 13 is connected to
the negative supply rail, causing T 2 to con-
duct via R 12 -D 7 -R 13 . The function of D; is
similar to that of Di as discussed earlier.
With all detector switches closed and the
loop unbroken, Ds conducts via T 2 and
Rh. Diode Ds does not conduct because its
cathode is connected to the positive sup-
ply rail via T 2 , as is its anode via Rs.
Capacitor O supplies a logic 1 to the
second input of bistable N 2 -N 3 after it has
been charged via Rs and R 9 . The two logic
1 s at the bistable inputs maintain a 1 at the
output, pin 4 of N 2 . As stated earlier, this
1 inhibits the sounding of an alarm.
Breaking the detector loop disconnects
R 13 from the negative supply rail, causing
T 2 to stop conducting. Its collector poten-
tial drops to nearly 0 V, so that Ds is for-
ward-biased via Rs and Rio. As a result, Ca
discharges in about 0.5 s via Rs, Ds and
Rio, its terminal voltage dropping to about
+1.8 V, which represents a 0.
The 0.5 s delay produced by Cs-
(Rs+Rni) assists in the prevention of false
alarms by interference spikes and other
transients in the loop circuit such as by
doors shaking in the wind.
Control terminal pin 1 of the bistable
accordingly changes from 1 to 0, so that
the level at the output terminal changes
from I to 0, where it will remain latched
in the absence of an alarm condition until
the other control terminal, pin 6 of N 2 ,
changes state, i.e., until the system is
switched off. The condition necessary for
the generation of alarm signals is a Oat the
output of the bistable.
Sounder timing
The alarm system has provision for two
sounders, one low-power internal alarm
such as an active piezo-electric buzzer,
and one high-volume external alarm such
as a 12 V bell.
The circuit automatically switches off
each of the alarm sounders after a reason-
able period of time: 4 minutes for the ex-
ternal sounder, and 8 minutes for the
internal sounder. The individual timing
circuits may be altered, however, to suit
personal preference.
Low-power CMOS timers Type 555
(IC 2 ) and 556 (IC 3 ) are used in the interest
of battery economy. When the circuit is
switched on, the timers are simultaneous-
ly released from the reset condition be-
cause their pins 4 are taken logic high.
Internal sounder
While the system is on, any break in the
detector loop, such as by a protected door
or window opening for longer than 0.5 s,
initiates operation of the internal sounder.
When the loop is broken, Cb passes the
l-to-0 transition at the output of the bi-
stable to pin 2 of IC 2 , which is triggered
into monostable operation for a period of
about 8 minutes. Network C 6 -R 16 forms a
differentiator to sharpen the trigger pulse.
On entering the premises, residents
have about 15 s to switch off the system
before the monostable switches on the in-
ternal sounder. Prior to the arrival of the
trigger pulse at pin 2 of MMV IC 2 , its out-
put, pin 3, is normally at 0. This level
keeps Tr off via base resistor Ri+ Immedi-
ately upon the arrival of the negative-
going trigger pulse at pin 2 of IC 2 , its out-
put rises from 0 to 1. This level is main-
tained for about 8 minutes as determined
by C 8 -R 17 . Transistor T 4 is switched on,
and actuates the internal sounder in its
collector circuit. When the 8-minute peri-
od has lapsed, the low level at pin 3 of IC 2
causes the internal sounder to be turned
off by T 4 .
For convenience of testing during the
construction and installation stages, LED
D« provides a visual indication of circuit
operation without the internal sounder
being connected. If actuated, the internal
sounder is switched off simultaneously
with the system.
External sounder
The operation of the external sounder cir-
cuit is slightly different from that of the
Fig. 2. Circuit diagram of the intruder alarm. Note that the timers, IC2 and IC3, must be
low-power versions to ensure minimum current drain from the battery.
12.34 elektor india decombor 1989
internal sounder. Assuming that the sys-
tem is switched on and the detector loop
not yet broken, the output of bistable N 2 -
Nj is at 1 . Capacitor Cs charges rapidly via
Ds, until its terminal voltage is also at 1.
Subsequently, Ts is turned off by the 0
supplied by inverter Nr. Timer IC 3 is not
yet triggered into action, so its output ter-
minal, pin 9, is at 0. Hence, darlington
transistor T 6 -T 7 is kept off in the absence
of an alarm signal — external sounder Bz 2
is not actuated.
Circuit ICs, a CMOS Type 556, contains
two timers Type 555. Pin 4 of the first 555
in the chip is held logic high via R 2 -D 1 -R 1 ,
so the timer is ready to be triggered. The
instant the detector loop is broken, the
l-to-0 pulse transition at the output of
bistable N2-N3 causes Ds to block, enab-
ling Cs to discharge through Ris. The time
constant formed by these two components
introduces a delay of about 15 s in the
transition from 1 to 0 at the input to inver-
ter Nr. After this delay, the resultant tran-
sition from 0 to 1 at the base of Ts causes
the transistor to conduct. The collector
voltage of Ts drops from 1 to 0, and the
negative-going pulse edge is differen-
tiated by Cio-R24to be passed as a sharp-
ened trigger pulse to pin 6 of dual timer
IC 3 . The first timer in IC 3 is configured as
a monostable with a 4-minute time period,
the output of which is used to control the
second timer circuit, which is configured
as an astable multivibrator (AMV). This
circuit can produce its 1-s on/off pulse
rate only during the 4-minute period of
operation set by C 11 -R 25 for the preceding
monostable in the IC. The time period, f,
in seconds can be calculated from
t = 1.1( Ci 1R25 )
Output pin 5 of the first timer is normally
at 0 until the arrival of an input trigger
pulse, whereupon the output state
changes abruptly from 0 to 1. Pin 5 is
wired to the reset input, pin 10, of the
second timer in the 1C package. When
taken high, this pin enables the AMV to
oscillate at a rate of 1 Hz during the 4-
minute period defined by the first timer.
The period (in seconds) of the oscillator
signal is calculated from
/ = 0.7Cm( R 26 + 2i?27 )
The square-wave oscillator signal drives
darlington transistor pair T<.-T7, so that the
external sounder, Bz2, is switched on and
off at a rate of about 1 s until the 4-minute
monostable period has lapsed. As with the
internal sounder, a visual indication of
external alarm activity is provided. Diode
D12 protects T? from transient voltage
spikes generated as the current through
the inductance formed by Bzz is inter-
rupted. Capacitors C12 and Cis are for de-
coupling and do not form part of the
timing circuits.
System assurance bleep
Provision has been made for a system as-
surance bleep to indicate that the system
is functional, prior to the resident's depar-
ture from the premises. Two assurance
bleeps are generated: one before the end
of the 15-s switch-on delay at the instant
of switch-on, and one as the exit door is
opened for departure.
While the system is switched off, C-i
has no voltage on it because Ti conducts.
Following switch-on, the 1 5-s delay before
the system becomes 'live' allows time for
the injection of a short control signal di-
rect to the internal sounder control tran-
sistor, T-t, bypassing timer IC 2 .
When the circuit is switched on, T 2 and
D 2 become non-conductive so that C 4 is
allowed to charge via Ris and R14 in about
0.25 s, which in effect momentarily causes
the base of T 3 to be taken low via R14. The
upshot is that both T 3 and Tt conduct just
long enough to enable the internal
sounder to produce a short bleep.
The same process occurs with T 2 and
Dh which, like D 2 , is connected to the junc-
tion of Cj and R14, except that in this case
the charging of C4 is initiated by the break-
ing of the detector loop w r hen a protected
door or window is opened.
Construction
A convenient and low-cost method of con-
struction is to use readily available copper
SRBP shipboard with 0.1 -inch hole spac-
ing. The use of sockets for the ICs is rec-
ommended, but the layout of components
is not at all critical.
Inter-component wiring is by thin in-
sulated wire. If stranded wire is used, care
must be taken to avoid unintended con-
tacts by loose unsoldered strands.
The external wires are connected to ter-
minal posts on the board. The two alarm-
test LEDs are purposely located on the
board for visual access during testing.
A separate box may be required to ac-
commodate the battery, and possibly the
mains adapter.
The on/off switch is either a key-oper-
ated type, or a cheaper standard on/ off
miniature toggle switch. A reasonable
compromise as regards safety might be to
use a standard SP5T toggle switch, and to
conceal it from view either complete with
the electronic assembly, or in a small sep-
arate enclosure.
Further practical
considerations
The door and window switches are mag-
netically operated types that have the ad-
vantage of not drawing current from the
battery. Constructors wishing to include a
motion detector of some sort in the loop
must bear in mind that such a device may
well draw 20 mA or more whether actu-
ated or not, which would have to be taken
into consideration when choosing the bat-
tery and the associated charger. Also, the
motion detector requires a separate cable
to carry its supply voltage. One approach
might be to replace the single-pole on/ off
switch with a double-pole (DPDT) type,
the other pole of which is used to connect
the +12 V to the motion detector only
while the system is switched on, assuming
that the battery is being recharged during
the off condition.
The cable-test loop shown in the circuit
diagram provides an indication in the
event of the loop having been tampered
with, for instance, cut by a prospective
intruder who plans a return visit while the
house is unoccupied. It would need to be
a separate pair but within a two-pair
cable; if both pairs are cut simultaneously,
the system would be switched on, and the
detector loop to be broken, so that the
alarm is set off immediately. If such a
situation is thought unlikely, the cable-
test loop may be omitted, and a substitute
wire link installed on the board. The de-
tector loop would then need to be twin
PVC insulated cable of, say, 7x0.2 mm
running from the board to each detector
in turn, and back to the board via the
unbroken wire of the pair.
The choice of the external sounder is
entirely up to the constructor, but care
should be taken not to overload the tran-
sistor driver or the battery. The author
used a weatherproofed sounder giving a
choice of continuous or warbling tone at a
sound level of 107 dBA for only 20 niA of
current drain from the 12 V battery. It is
standard practice to enclose the external
sounder in a weatherproof enclosure, in-
stalled high up on the wall out of easy
reach, and with its supply cable hidden
behind the box as an anti-tamper precau-
tion.
elektor indie december 1989 12.35
PROTECTING ASYNCHRONOUS MOTORS
by Mehrdad Rostami, University of Tehran, Iran
The circuii described here was designed for
protecting heavy-duty asynchronous motors
during the start-up period. As is well-known,
without protection such motors may easily
get damaged by poor starting. The circuit
may also be used lor other applications
where a trip circuit needs to be triggered,
such as, for instance, in the monitoring of
liquid levels.
Every motor has a time-speed character-
istic that shows how, or otherwise, it starts
and reaches its normal speed. A number of
such curves are illustrated in Fig. I. If the
characteristic of a particular motor is similar
to the lower (bold) one. any attempt at start-
ing the motor should be stopped immediate-
ly and the motor in-spected thoroughly. The
dashed curve in-dicates the lower limits of
acceptable motor performance, while the
upper curve shows normal values of a prop-
erly functioning motor.
Circuit description
The circuit diagram in Fig. 4 consists of five
identifiable blocks: (1) oscillator and time
base — IC4, IC5 and IC6; (2) address unit
and memory — IC7. ICS and IC9: (3) shaft
pulse receiver and counter — ICI4 and IC15;
(4)comparator — IC12 and IC13: and (5)
automatic stop unit — FF1 and FF2.
The input to the circuii consists of pulses
generated by a rotary encoder comprising an
opto-coupler and perforated man-made fibre
disk fitted securely on to the shaft of the
motor as shown in Fig. 2. The pulses gener-
ated by the opto-coupler are applied to
receiver/counter ICI4 and then to counter
1CI5.
The 555 oscillator.. IC4. generates 50 Hz
pulses that are divided by 5 in IC5. The out-
put of this 1C is taken to switch S 1 and also
applied to a second :5 divider. IC6.
The output of either divider may be
selected by SI and from there applied to cas-
caded circuits IC9 and IC 10. The output of
1C 10 is used to reset the shaft pulse counters,
IC 1 4 and IC 15. at the end of each period of
0.5 s or 0. 1 s depending on the setting of S 1 ,
and also to clock the address unit, IC7 and
IC8.
The eprom must be loaded with the data
of the appropriate motor curve. If, for
instyance, the rotary encoder is supposed to
send eight pulses in the first 0.5 s period (SI
set to 2 Hz) — which, of course, depends not
only on the rotary speed of the shaft of the
motor, but also on the number of perfora-
tions in the disk — the first memory cell of
IC 1 1 must be loaded with 00001000. The
number of pulses is determined from the
timing diagram of the relevant motor: a typi-
cal ttime vs rotary speed characteristic is
shown in Fig. 3.
Similarly, if the pulse generator is sup-
posed to send 12 pulses in the second 0.5 s
period (SI set to 2Hz). the second memory
cell of the eprom must be loaded with
00001 100. This process must be repeated for
each subsequent 0.5 s period (up to a total of
20 seconds, when a properly working motor
will have started).
The outputs of the eprom and the shaft
pulse counters are applied to two Type 7485
comparators, 1C 1 2 and IC 1 3.
At the end of each 0.5 s period. IC9 gen-
erates a pulse that is used to drive one of the
inputs of and gate N2 high. When the level
at pin 7 of comparator IC 1 3 is also high, the
second input of N2 goes high. also. This
results in the output of this gate becoming a
logic 1 . which is applied tots of and gate N3.
The second input of N3 is supplied by
autostop unit FFI, a D-type bistable. This
bistable is reset by and gate N I when
address 00010100 is applied to the eprom.
Its Q output then goes high, which causes
the second input, and thus the output, of gate
N3 to go high. This causes a second D-type
bistable, FF2, to be set. When that happens,
the coil of a trip device in the starting circuit
of the motor is energized so that the starting
circuit is broken.
Circuits IC 1 2 and IC 1 3 compare the data
input from the eprom with that from coun-
ters 1C 1 4 and IC15. If these data streams arc
identical, pin 7 of IC 1 3 re-mains low', pre-
venting the operation of the automatic stop
unit.
Schmitt triggers N4. N5 and N6 form an
auto reset circuit for setting/resetting the
bistables and returning the counters to their
original state.
Fig. 1. Time-speed characteristics of a an asyn-
chronous motor. The lower (bold) curve indi-
cates a defect motor; the dashed curve indi-
cates the lower limit of acceptable performance;
and the upper curve is typical for a properly
functioning motor.
Fig. 2. The rotary encoder consists of an opto-
coupler and a perforated man-made fibre disk
fitted on to the shaft of the motor
Fig. 3. Typical time vs rotary speed diagram of
an asynchronous motor. A properly working
motor should start within 20 seconds.
12.36 Glektor india december
Fig. 4. Circuit diagram of the protection unit.
\l \\ PRODUCTS
Digital V-A-F Meter
Jivan has introduced Digital V-A-F
meter. This measure voltage. Ampere &
Line Frequency. It has a compact size of
96 (H) x 96 (W) x 170 (D) mm. It directly
measures measures voltage upto 600 V,
with P.T., it can measure any desired
voltage. The current range is upto 10 A,
but with C.T. , it can measure any desired
current. The frequency range is from
20.000 to 99.99 Hz.
Jivan Electro Instruments • 394, GIDC
Estate • Makarpura • Baroda-390 010 •
POWER CONNECTORS
G.H. INDUSTRIES introduces Power
Connectors Pitch 3.96 mm, 5.08 mm,
5.0-7. 5 mm, 5.0 mm Range : 2 Way to 22
Way.
Male Square Pin Headers are also availa-
ble with or without Friction Lock,
Straight or Right Angle. Current - Rat-
ing 7.0 Amps. Voltage 250 Volts.
ISP-3960
G. H. INDUSTRIES • 84-B Government
Industrial Estate • Kandivli • Bombay-
400 067 •
elektor india december 1989 1 2.37
SCIENCE & TECHNOLOGY
Intelligence, Intentionally and Self Awareness
by Dr T. Farrimond, University of Waikato, New Zealand
This paper deals with some of the problems in ascribing
intelligence to computers. It is suggested that machines which only
process the symbols of language are not intelligent even though they
may produce an output similar to that from an intelligent human.
It is maintained that self awareness in humans, coupled with the ability
to interact directly with the environment by means of the senses,
is central to intelligent functioning, which includes the development
of a social/ecological conscience.
In his article “Artificial Intelligence”, M.
Seymour 1 provides an interesting and infor-
mative account of some of the problems met
by computer designers in attempting to pro-
duce machines that exhibit artificial intelli-
gence. The article discusses arguments for
and against what constitutes artificial intelli-
gence including the existence or otherwise
of intentionality (Searle, 1 984)-. The present
paper examines some of the concepts from
the point of view of a psychologist, who was
a student at Manchester when Alan Turing
was working on the theoretical aspects of
information processing. The power of elec-
tronic devices has increased enormously
since that time, but perhaps there has not
been a similar growth in defining the termi-
nology used to describe computer activities
and brain activities.
At the simplest level there has been revival
of anthropomorphism, a condemnatory
appellation feared by biologists accused of
reading human characteristics into the
behaviour pattern of lower animals. Howev-
er, equally imprecise use of language is
exemplified by phrases such as ‘computers
talking to each other’. This is largely a mat-
ter of economy in the use of words, since it
is easier to use concepts already in existence
than to invent new ones, but there are dan-
gers in over-extending the concepts to
include things that are not justifiable. The
problem is that with terms such as intention-
ality it is difficult to provide a definition that
does not also include or imply the term
intention, which then also has to be defined.
In describing a spiral staircase, it is easier to
make a visual representation by drawing one
(or to wave one's arm to illustrate the con-
cept) than it is to describe it verbally. If this
is true for a concrete example such as this,
then for abstract concepts the difficulties
involved in using words to define them are
enormous.
Is the term intentionality sufficient to cover
those things the brain does that are different
from a computer? How does one recognize
intentionality? Can intentionality be proven
and is it important to do so? The concept of
intentionality is essential in dealing with
human affairs, particularly when legal dis-
putes arise and require resolution. We resort
to a court of law where proof of intention
may well determine the outcome of a case.
Did the accused know what she was doing
when she set fire to her husband's bed? Evi-
dence may be produced to prove diminished
responsibility; a person may be described as
intellectually sub-normal and so not account-
able for his/her actions. The implications in
this case may be that the accused did not
properly understand that the outcome of the
action might be injury or death. Similar inca-
pacity may also be ascribed to a person
under the influence of drugs or suffering
from some mental disorder. The question of
responsibility is the key to determining
whether the sentence should be 10 years or
alternatively some form of medical treat-
ment. In each case what is examined are the
following.
(a) Could the individual predict the out-
come of the act that caused the accident
(is there an ability to follow a logical
sequence of events on a probabilistic
basis to a conclusion or variety of possi-
ble conclusions)?
(b) Did the person intend to set in train the
causal events that resulted in harm? If a
person accidentally backs into a lever
that releases a winch carrying a load of
iron, causing it to fall and kill someone it
is not the ability to understand the causal
relationships that determines guilt - but
whether there was intention to do harm.
In this example there was not.
(c) Was there an awareness on the part of
the accused that he or she was carrying
out the action?
Point (c) is releva., t, for example, in the case
of hypnosis. A woman under hypnosis may
be persuaded to rote-play the part of some-
one in authority and perform an act not nor-
mally acceptable to her simply because she
regarded herself as another person during the
period of hypnosis. In this case a causal
sequence of events has occurred in which
there is intention on the part of the subject
to carry out an act, but because self aware-
ness is absent, the individual would not be
regarded as culpable in law. Even though her
behaviour incorporates the two elements
usually considered necessary for intelligent
behaviour, i.e., it exhibits appreciation of
causality and also intentionality, she is not
seen as responsible for her behaviour. It may
be argued that intelligent human behaviour
involves these elements - causality, intention
and self awareness and for a computer to be
regarded as intelligent it also should exhibit
the same properties.
It is this point of self awareness which I con-
tend is different from intentionality and is
possibly the central issue in determining
whether behaviour is intelligent or not. It is
assumed that the use of the term intelligence
is a reference to human mental and
behavioural processes since these are the
only points of reference we have for what
we mean by intelligent behaviour.
External behaviour
Would a machine designed to look and move
exactly like a human being so that it would
12.38 eleklor india december 1989
be accepted at a barbecue (or even a social
function!) really be intelligent? One could
forgive the hostess for assuming that it is,
since from the outside the machine does all
the things normally expected of a human: it
speaks, moves about, listens attentively and
even laughs in the appropriate places.
It is tempting to argue that it is only the
behaviour of the machine that is important,
i.e.. outside appearances and behaviour are
all that matter. If these are indistinguishable
from human behaviour, the machine should
be regarded as human, and therefore intelli-
gent. Indeed, this may be the effect on the
hostess until it is demonstrated to her that a
group of electronics enthusiasts have con-
structed the machine and are operating it
remotely: one controlling locomotion, anoth-
er speech, and so on. Thereafter, the hostess
would no longer accept as fact that because
someone (thing) exhibits intelligent human
behaviour it is genuinely intelligent. This
emphasizes the problem that without further
knowledge about the controlling mecha-
nisms it is difficult to prove that a behaviour
pattern is intelligent or not. But is is obvi-
ously not safe to infer intelligence on
behaviour alone. In the example given, the
intelligence is elsewhere and is external to
the machine.
A distinction should be made between the
analogous behaviour and identical behav-
iour. Herein lies the distinction between
machines at present and humans. The
behaviour of a machine may be analogous to
that of a human without necessarily being
identical.
Although it may be the expressed aim of
engineers to produce intelligent machines, it
is doubtful whether they would want them to
be intelligent in the human sense, since they
may no longer wish to co-operate with the
inventor - and may prefer to go on strike.
Certainly, any organism (biological or
mechanical) with self awareness would also
be aware of its rights as a thinking being and
its utility as a tool (that is, slave) would be
reduced. An interesting prospect also opens
up in the area of culpability for mistakes. If a
machine is regarded as culpable and it trans-
gresses, what should its punishment be?
Absence of need for
programming
It has been envisaged that one day it may be
possible to build a machine that can think,
that is, need not be programmed to perform
its functions. This statement as it stands per-
haps needs elaboration before its implica-
tions can be considered. If the term ‘thinks’
refers to performing certain analytical func-
tions, the similarity to human thinking is
restricted to one level of activity. It would be
necessary to define the term in other ways if
it were to include intentionality and self
awareness. The presence of one level of
functioning does not automatically mean that
the other levels are present. Terms such as
intelligence, cognition, perception, etc., have
evolved from attempts to categorize (by
using symbols) certain aspects of human
behaviour. The words are not specific but
incorporate implied connections with all
other aspects of human mental activity.
Gregory in his book The Intelligent Eye
emphasizes the relationships that exist be-
tween the eye and the brain. The eye is an
extension of the brain in a psychological as
well as in an anatomical sense. The unitary
nature of perception, cognition, intelligence,
etc., makes it difficult to talk about simply
one aspect of human behaviour without
automatically including all the others. It
would make little sense to examine human
cognition without at the same time consider-
ing intelligence, memory store, and percep-
tual abilities, for cognition depends upon
them all. Also, an individual's cognitive state
is constantly changing, not only from new
experiences, but by re-analysis of stored
information from within, where models exist
of the world (imagery) available to the indi-
vidual for the process of thinking, research-
ing and creating.
The capacity of the brain
In an attempt to duplicate the equivalent of a
neural net system as found in the brain,
experimenters have constructed electronic
networks with a large number of intercon-
nections. However, the human brain is not
simply a neural network. The complex of 10
billion (10’) interconnected brain cells con-
fers only one part of the brain's processing
power, for along with nerve cells there are
over five times as many smaller glia cells.
All these cells have numerous fine branches
extending from them to form interconnec-
tions with other nerve cells: some individual
cells may have several hundred connections,
others several thousand and in the cerebel-
lum certain cells may have one hundred
thousand connections. The number of inter-
connections has been estimated to be of the
order of 50 trillion (50 x 10”). Nor is this the
whole story. Memory storage in the brain
seems to involve changes in the protein
molecules associated with the nerve cells.
Additionally, certain glia cells are not fixed
relative to adjacent brain cells but may move
into active areas of the brain, thus modifying
the brain's structure in response to incoming
stimuli. Glia cells, unlike larger brain cells,
have the ability to subdivide as well as
move, so that their number and distribution
may change depending upon the activities of
the brain.
What makes the human brain so interesting
is that the owner is, to some extent, able to
observe his/her mental states and decide
upon a course of action thereby. This course
of action is not unchangeable but open to
modification. Even though humans have
characteristic patterns of behaviour by which
they may be recognized as individuals, it is
still possible for a person to examine past
behaviours and bring about a change for no
other reason than that a change is regarded
as desirable. This capacity makes human
behaviour notoriously unpredictable even
when we know a person very well. This is
not the same as Turing's 3 suggested incorpo-
ration into a machine of a ‘random element’
consisting of a random number series
which produces changes in the behaviour
of the machine. In human terms, such a
random element would be more character-
istic of psychotic human behaviour, where
there may be an absence of awareness of
the behaviour on the part of the psychotic
and little appreciation of its effect upon
others. Self awareness is the ability that
gives humans the capacity for controlled
variability and includes intentionality and
appreciation of causality.
The origin of self awareness
Although it is difficult to be specific on
this point since we no longer remember
what we experienced in the few months
preceding our birth, it is possible to con-
jecture that our sense of ‘self’ begins to
develop quite some time before birth.
Acoustic images of developing foetuses
show them yawning, moving, sucking
their thumbs, etc., indicating the presence
of kinaesthetic and tactile awareness.
There seems little doubt that, like Tristram
Shandy, we are responding to, and becom-
ing aware of, our own bodies in relation to
the environment surrounding us. In other
w'ords, we are developing self awareness.
Self awareness includes the development
of body image, that is, the knowledge that
our bodies are unique, yielding sensations
that are related to each other. Visual and
tactile investigation by a young baby of its
body yields a complex integrated pattern
of sensations that, in conjunction with
kinaesthetic feedback from muscles and
joints, gives the child a sense of personal
identity that is different from all other
objects in the environment: other objects
are regarded as external to the self. To
achieve this development of body image,
the child must move relative to the envi-
ronment. so that it experiences variations
in the size of objects as distance changes
and variations in shape as viewing angles
change. Both the distance information
gathering senses of vision and hearing are
co-ordinated with the body senses of
touch, pressure, pain, temperature and
kinaesthetic feedback, to produce an orga-
nized pattern of information resulting in
elektor india december 1989 1 2.39
self awareness.
The experiment by Held and Hein (1963) 4
with kittens indicates that visual ability
requires integration of changing visual
patterns (brought about by moving in the
environment) with simultaneous stimula-
tion of body senses and locomotor activity
on the part of the animal. In this experi-
ment, two kittens were kept in the dark
until their eyes opened. Then they were
placed at opposite ends of a bar pivoted at
its centre so that it could rotate. Only one
kitten, 'a', had its feet on the floor and so
could walk around in a circle. It could also
turn around on the spot because of the
design of the apparatus. The other kitten,
'B\ stood in a basket that prevented foot
contact with the floor but, because of an
interlinking system of gears and chains, it
was moved whenever kitten a moved: it
could not initiate movement itself. Both
kittens therefore received similar visual
stimulation. When the kittens were
released after 30 hours, kitten a could
make normal visual responses, such as
avoiding a cliff, blinking to avoid an
object approaching the eye and avoid
obstacles. Kitten b was unable to do any of
these tasks and only learned to see when
allowed to walk.
It has been stated that “artifical intelli-
gence is the study of computer programs"
(Boden) 5 . In humans, it would perhaps be
more accurate to say that intelligence is a
function of the body and equates with sen-
sitvity to external and internal stimuli. The
new born baby has no program derived
from outside sources, although it shows
responses: exhibiting sensitivity to (and
reflex movements away from) painful
stimuli. Light and sound convey little
meaning at this stage; learning is initially
related to the body senses. For example, if
the baby makes random movements of the
hands, it may strike the side of the cot and
receive a sensation in that hand. If the
baby strikes its own face, it receives a sen-
sation in the face as well as in the hand.
This is a unique experience different from
all other contacts with the world outside
the individual's body. The baby soon asso-
ciates these sensory inputs with the inter-
nally derived sensations from the muscles
that are involved in making the move-
ments, so from the beginning sensory
information establishes a complex body
image. This is later extended to include
visual and auditory patterns and rapid
learning occurs. It is worth noting that lan-
guage need not be involved. A deaf child
exhibits intelligent behaviour solely by
observation of the environment: recogni-
tion of a person's facial expressions or ges-
tures is an early form of communication.
In humans, simple signals and signs later
become more complex to include written
and verbal symbolization so developing
into language as used in the conventional
sense. It is at this level of symbolization
that it becomes possible to manipulate
words or numbers as models of the envi-
ronment. The usefulness of the scientific
method has depended upon establishing an
accurate correspondence between symbols
and reality. When the symbols no longer
do the job of predicting or explaining, one
returns to the experiment as exemplified
by Faraday 6 .
There is a danger that the symbols may be
regarded as the repository of intelligence,
when in fact the symbols only exist
because of Ihe intelligence used to con-
struct them initially. Mechanical manipula-
tion of symbols according to the rules of
language may bring benefits in solving
problems, but the program responsible for
the manipulation (itself a language) lacks
the attributes of self awareness and sensi-
tivity to the environment that characterize
human intelligence.
Brain and machine
translations
A machine may reproduce functions that
may be similar to human ones, for in-
stance, translating English into French.
The process of translation is established by
comparison of the two sets of visual sym-
bols, since the languages follow very simi-
lar patterns. Languages describe the vari-
ables in our environment and these are, in
most physical aspects, common to all
societies. The same things are given differ-
ent symbols (either auditory in the case of
speech or visual in the case of written lan-
guage). The dynamics of events in the
environment are also constant: 'a girl
runs', 'an object falls', 'a goat jumps', and
so on. Therefore, translation involves
matching two symbolic patterns, but to
produce language, a perceptive organism
must first observe the environment and
establish a linguistic model of the ‘real
world', which may be used for interchange
of ideas. In the case of a second language,
some important similarities are estab-
lished, for instance, finding what symbols
in French stand for man, woman, girl,
goat, etc., after which translation is rela-
tively easy because of the communality of
experience of the environment embodied
in human languages. The translation of
Egyptian hieroglyphics was not possible
until the discovery of the Rosetta stone
where the same message had been record-
ed in hieroglyphics, Greek and Coptic.
The recognition of the name of Ptolemy,
which occurred in all these versions, made
it possible for Champollion to equate the
unknown hieroglyphics with a known lan-
guage and so produce a translation. Lan-
guages have contained within them a
causal pattern echoing the environment
from which the language was derived.
The interesting aspect of languages is that
once they have been established, they may
be processed in a variety of different ways
because of the built-in degree of corre-
spondence to our world, which makes
them useful tools. However, language can
not express unambiguously all aspects of
the real world since linguistic concepts of
language (including mathematics) relate to
generalities and not specifics. Linguistic
devices may be used to define a particular
dog as spaniel, but specificity requires
more descriptive information. We soon
reach a point where language is no longer
capable of conveying the information that
a few seconds' direct contact with the dog
would provide. State of health, condition
of coat, friendly or not, does it like you,
how old is it, how heavy, etc. Language is
a substitute for reality, and this limitation
extends to all descriptive applications.
The problem of ascribing intelligence to a
device that solely processes language is
revealed if a nonsense-language is used.
The machine may produce 'solutions’ to
nonsense problems fed into it (following a
set of rules), but these would be meaning-
less. The machine is no less capable than
machines using real language, nor is its
program less complex. The only difference
between a nonsense machine and a lan-
guage processing machine is the degree of
correspondence of the symbols used to our
environment and this is something that an
external observer perceives. This is intelli-
gence by implication, that is, Ihe recogni-
tion that certain activities resemble (or dif-
fer from) human intelligence: in the case
of language processing, intelligence is a
function of neither the machine nor the
program.
If a black box processes problems, it is
tempting to regard the machine (or pro-
gram) as intelligent since its behaviour
resembles that of intelligent humans. If the
black box is enlarged to make a room
capable of housing hundreds of thousands
of people, these may be arranged to pro-
cess information in the same way as a
machine. Chains of individuals handle the
input, make available stored information
and present an output as a machine does.
In this case, where does the intelligence
lie? The grouping of individuals is analo-
gous to the circuitry of a machine, but no
'group intelligence’ is generated simply by
the use of a number of individuals. The
instructions to the subjects are carried out
by the occupants of the room, but each
person is simply carrying out part-func-
tions, the implications of which are not
recognizable since their relation to other
functions is not apparent. The program
12.40 elektor india december 1989
represents the instructions that the workers
are carrying out. Intelligent performance is
recognizable only by observing thg perfor-
mance of the whole group. Intelligence
then is not in the program itself, but in the
way the program was designed. This sug-
gests that it is possible to design a machine
that performs according to its program-
ming in an apparently intelligent way
without it necessarily being intelligent.
The machine would need to organize its
behaviour by itself, monitor the environ-
ment and be responsive to it and be aware
that it was doing so if its behaviour were
to be equated with human intelligence.
Intelligence
A definition used by Alfred Binet in-
volves at least four factors:
1 . Direction - the ability to set up a goal
and work toward it:
2. Adaptability - the ability to adapt
onself to the problem and use
appropriate means to solve it;
3. Comprehension - the ability to
understand the problem;
4. Self evaluation - the ability to
evaluate one's performance and to
determine the correctness of approach.
Examples of intelligence in humans cover
an enormous number of activities ranging
from simple identification of objects to
solving complex problems involving the
practical manipulation of equipment and
the development of theoretical models
(based on the result of experimentation).
This involves both language and mathe-
matics.
In Binet's factor of self-evaluation, the
concept of self awareness is implicit since
to evaluate one's own performance
requires that one must be aware of what
the performance was, who the performer
was and that the evaluator of the perfor-
mance was the original performer. This
type of self-analysis with its recognition of
individual identity is a fundamental fea-
ture of human intelligent behaviour. Occa-
sionally one finds in the literature refer-
ence to ‘idiots savants'. Really, the term is
self-contradictory since idiocy and sagaci-
ty are mutually incompatible. The term is
used to describe those individuals who,
while showing limited general intelli-
gence, are somehow able to perform bril-
liantly in a specific area, for example,
adding up large columns of figures, or
working out the day on which a particular
dale falls in the calendar fifty years hence,
etc. In human terms, they would not be
regarded as intelligent but rather as having
a processing facility for certain data.
Wechsler 7 described intelligence as the
purposeful and rational ability to deal with
the environment. Human intelligence
requires that an individual be able to inter-
act with the environment, perceive rela-
tionships, predict events and be aware of
the effect of his/her actions on others. This
is an example of primary intelligence.
Symbolic representations in the form of
language and mathematics are evolved
later as convenient tools for processing
information derived from primary intelli-
gence. As stated earlier, when symbolic
systems have been constructed, these lend
themselves to processing in a variety of
different ways, but they are the outcome of
intelligence rather than intelligence per se.
Terms such as cognitive science or artifi-
cal intelligence as applied to the process-
ing of symbols refer to aspects of human
abilities and there is a danger in attributing
too much to processing functions solely on
the grounds that they reflect some aspects
of human intelligence.
In the introduction to his book on inten-
tionality, Searle has argued for the inclu-
sion of mental activities when concepts
such as intentionality are considered; he
rejects "any form of behaviourism or func-
tionalism, including Turing machine fun-
actionalism, that ends up by denying the
specifically mental properties of mental
phenomena”. My own thoughts from a
psychological viewpoint also stress cau-
tion in reading too much into machine per-
formance, since there is a danger of estab-
lishing a form of anthropomorphism that
may militate against exploration of human
brain functions by model making.
Systems of linguistic analysis and
response are closed systems (at present).
Once the rules are provided, behaviour is
determined by the logic of the system,
even though changes in patterns may be
affected by the introduction of new data.
Self awareness would represent a constant
monitoring by the system of its perfor-
mance in relation to the world outside and
to itself. Some aspects of social self
awareness are outlined by Duval and
Wicklund (1972), Argyle (1969) and
Fcnigstein, Scheierand Buss (1975) 8 .
Elements introduced by human self aware-
ness are not necessarily logical or related
to a predetermined goal of efficiency or
accuracy. Departures from a logical path
may be brought about by the recognition
of similarity between the 'individual' and
other individuals (which is the beginning
of social intelligence and moral responsi-
bility). Emotions such as pity, compassion,
love, etc., may produce departures from a
logical behaviour pattern since self aware-
ness links all forms of behaviours with
oneself. Ethical considerations involving
feelings of empathy for others arise,
involving both animals and humans. ‘If I
were a gorilla, would I like my habitat
destroyed?’, and so on. Introspection
brings a new level of internal control of
behaviours that may seem unintelligent
(when in love, for instance), yet each
behaviour is intelligent within the frame-
work of the individual's perception of
his/her feelings. The list of human
attributes that may influence intelligent
behaviour is enormous and includes, along
with love, altruism, self-sacrifice, admira-
tion, aesthetic appreciation, and so on.
Without such sensitivity to environmental
factors, it would be difficult to argue that
intelligence was at work. The current con-
flict between developers and conservation-
ists is an outcome of a wider intelligence
coming into conflict' with commercial
intelligence. It would seem prudent from
the outset that exploration into the areas of
cognitive science and artificial intelligence
should not be restricted to a narrow spec-
trum, but should attempt to deal with the
wider issues involved in intelligent
behaviour.
References
1. “Artificial Intelligence”, M. Seymour,
Elektor Electronics India, June 1988.
2. Intentionality: an Essay in the Philoso-
phy of Mind, John R. Searle, CUP 1983.
3. Alan M. Turing, p. 133, Sara Turing,
W. Hel ler & Sons. 1959.
4. “Movement produced stimulation in
the development of visually guided
behaviour" R. Held and A. Hein. Journal
of Comparative and Physiological Psy-
chology, 56. 872—876. 1963.
5. Artifical Intelligence and Natural Man.
p. 3, Margaret Boden, The Harvester
Press, Hassocks.
6. Michael Faraday: a Biography, L.
Pearce Williams. Chapman and Hall.
1965.
7. The Measurement and Appraisal of
Adult Intelligence, D. Wechsler. Williams
and Wilkins (1958), Baltimore Md.
“Intelligence defined and undefined: a rel-
ativistic appraisal”, D. Wechsler, Ameri-
can Psychol., 30, 135-139 (1975).
8. “Public and Private Self Conscious-
ness: Assessment and Theory'', A. Fenig-
stein, M.F. Scheier and A.H. Buss, Jour-
nal of Consulting and Clinical
Psychology, 43, 4, 522-527 (1975)..
Social Interaction, M. Argyle, Atherton
Press (1969), New York.
A Theory of Objective Self Awareness, S.
Duval and R.A. Wicklund, Academic
Press (lt>72) New YHork.
elektor india december 1989 12.41
DC-DC POWER CONVERTER
T. Wigmore
This high-efficiency step-up converter supplies up to 30 V at 75 W
when powereo from a 12 V car battery. The converter is ideal for
many mobile and other out-of-doors applications: it functions as a
power source for your DC-operated soldering iron, RF power
amplifier, or NiCd battery charger for portable equipment such as a
flasher or a video camera.
DC-DC converters for stepping up the car,
battery voltage are generally based on a
switched-mode power supply (SMPSU)
or a power multivibrator driving a trans-
former. The power converter described
here is based on the first principle, and
uses the Type TL497A integrated circuit
from Texas Instruments. This device en-
ables good voltage regulation with low
output noise to be achieved fairly easily,
and in addition guarantees a relatively
high conversion efficiency.
Design background
The converter described is of the flyback
type. The flyback principle is the only
practical way of generating a direct out-
put voltage from a lower direct input volt-
age.
The central switching element in the
converter is power S1PMOS transistor Ti
(see Fig. 1 ). When it conducts, the current
through Li rises linearly with time. Dur-
ing the on-time, magnetic energy is stored
• Flyback-type step-up converter
• no special inductor required
• input voltage: 12 VDC
• output voltage adjustable between 20
and 30 V
• maximum output power: 75 W
• efficiency: 70%, independent of load cur-
rent
• voltage reduction at load variation from
zero to maximum: <200,mV
• ripple voltage: <500 mVpp.
in the inductor. The moment the transistor
is turned off, the inductor functions as a
source of magnetic energy, which is sup-
plied as an electric current to the load via
Di. In this process, it is important that the
transistor remains off during the time
taken by the magnetic field to decay to
zero. When this condition is not met, the
current through the inductor rises to the
saturation level. An avalanche effect then
causes the current to increase very rapid-
ly. The relative on-time, or duty factor, of
the transistor control signal must, there-
fore, not be allowed to reach the value of
one.
The highest permissible duty factor is
dependent, among other factors, on the
output voltage, because this determines
the rate of decay of the magnetic field
strength. The maximum output power
that can be supplied by the converter is
governed by the maximum permissible
peak current through the inductor, and
the frequency of the switching signal. The
limiting factors here are mainly the satu-
ration instant and the maximum tolerable
ratings for the copper losses in the induc-
tor, and the peak current through the
switching transistor (remember that
a'burst' of a particular energy content is
supplied to tire output at each switching
period).
TL497A
The operation of this integrated circuit is
rather unconventional, so that a brief de-
scription is given below.
In contrast to widely used fixed fre-
quency, variable duty-factor SMPSU con-
troller ICs, the TL497A is qualified as a
fixed on-time, variable frequency device.
This means that the duty factor is control-
led by means of frequency variation to
maintain a constant output voltage. This
method results in a fairly simple circuit,
but has the disadvantage of the switching
frequency reaching down into the audible
range when the load current is low. In
actual fact, the switching frequency
becomes lower than 1 Hz when the con-
verter is not loaded. The slow ticks heard
as a result are the charge pulses applied to
the output capacitors to maintain a con-
stant output voltage. In the absence of a
load, the output capacitors are, of course,
slowly discharged by the voltage sensing
resistors.
The on-time of the oscillator on board
the TL497A is fixed, and determined by
Ct. The oscillator may be disabled in three
ways: first, if the voltage at pin 1 exceeds
the reference voltage (1 .2 V); second, if the
current through the inductor exceeds a
certain maximum; and third, via the in-
Fig. 1. Circuit diagram of the step-up converter.
hibit input (this is not used here).
During normal operation, the oscilla-
tor causes Ti to conduct so that the induc-
tor current rises linearly. When Ti is
switched off, the magnetic energy stored
in the inductor is used to charge the out-
put capacitors. The output voltage, and
with it the voltage at pin 1 of the TL497A,
rises a little, so that the oscillator is dis-
abled until the output voltage has
dropped to a sufficiently low level. This
process is repeated cyclically, at least, in
theory.
In a configuration with real compo-
nents, however, the voltage rise caused by
the charging of the capacitors within one
oscillator period is so small that the oscil-
lator remains enabled until the inductor
current reaches the maximum value
defined with R 2 and IU (the voltage drop
across R 2 and R 3 is 0.7 V at this stage). The
current rises in steps as shown in Fig. 2b
because the duty factor of the oscillator
signal is greater than 0.5.
When the maximum current is
reached, the oscillator is disabled, and the
inductor is allowed to pass its energy to
the capacitors. In this condition, the out-
put voltage rises to a level high enough to
keep the oscillator disabled via pin 1. The
output voltage drops, and a new charge
cycle commences.
Unfortunately, the switching oper-
ations outlined above are coupled to rela-
tively high losses. In a practical
application, this problem is resolved by
making the on-time (i.e., Ci) large enough
to ensure that the inductor current does
reach the maximum within a single oscil-
lator period (see Fig. 3). The solution in
this case is the use of an air-cored induc-
tor, which has a relatively low self-induct-
ance.
Some waveforms
The timing diagrams in Fig. 3 show the
signal waveforms at the main points in the
circuit. The central oscillator in the
TL497A operates at a low frequency
(lower than 1 Hz if the converter is not
loaded). The switch-on instant, shown as
the rectangular pulse in Fig. 3a, is deter-
mined by capacitor Ci. The switch-off
time is determined by the load current.
During the on-time, Ti conducts so that
the inductor current rises (Fig. 3b). In the
non-conductive period after the current
pulse, the inductor functions as a current
source. The TL497A compares the attenu-
ated output voltage at pin I with its inter-
nal reference voltage of 1.2 V. If the
measured voltage is smaller than the ref-
erence voltage, Ti is driven hard again to
enable the inductor to store energy .
The above charge and discharge cycles
cause some ripple voltage on the output
capacitors (Fig. 3c). The feedback arrange-
ment enables the oscillator frequency to
be adjusted for optimum compensation of
voltage losses caused by the load current.
The timing diagram in Fig. 3d shows
considerable swing of the drain voltage
owing to the relatively high Q (quality)
factor of the inductor. Although the para-
sitic oscillations do not affect the normal
operation of the power converter, they
may be damped with the aid of a 1 k(2
resistor in parallel with the inductor.
From theory to practice
Naturally, a switch-mode power supply is
designed for maximum rather than quies-
cent output current. High efficiency and a
stable output voltage with little ripple are
also prime design goals.
In general, the load regulation charac-
teristics of a flyback type switch-mode
power supply give little cause for concern.
During every cycle, the on/ off ratio is ad-
justed in accordance with the load cur-
rent, so that the output voltage remains
fairly stable in spite of large load current
variations.
The situation looks a little different as
far as the overall efficiency is concerned.
A step-up converter of the flyback type
typically generates relatively large cur-
rent surges, which cause considerable
power losses (remember that power rises
exponentially with current). In practice,
however, the proposed converter has a
total efficiency higher than 70% at maxi-
mum output current, which is remarkable
given the simplicity of the design.
Fig. 2. Showing how the inductor energy is built up under the control of the oscillator
signal.
elektor india december 1989 1 2.43
a ut
tin
I 'out
oscillator
n
*
1 1
b 1
flyback period 890030 • 12
Fig. 3. Timing diagrams of the main signals in the circuit. The current reaches its maxi-
mum value within one period of the oscillator signal.
The switching frequency at maximum
load is made as high as possible to allow
the use of a relatively small self-induct-
ance. The practical circuit is based on an
air-cored inductor. Significant losses
caused by a ferrite core are thus avoided.
A fast power-FET of the S1PMOS type
is used to switch the inductor current. The
Type BUZ10 or BUZ10A was chosen be-
cause of its short recovery time. To
achieve acceptable efficiency, the transis-
tor must be used as a switching element.
Parts list
Resistors (±5%);
Ri = IkO
R2;R3 - 0£11 ; 4 W
R« = 18K11
Rs = 1 K2
Pi = 10U2 preset H
Capacitors:
C: = 680p
C2IC3 = 470g; 35 V; radial
C4 = 1000g; 16 V; radial
Inductor:
Li = 30 pH (home-made, see text)
Semiconductors:
Di = BYV79
Ti= BUZIOor BUZ10A
ICi = TL497A
Miscellaneous:
Heat-sink for Ti.
PCB Type 890030
12.44 elektor india december 1989
This, in turn, requires it to be driven into
saturation, resulting in a relatively long
turn-off time. Obviously, the longer it
takes for the transistor to interrupt the
inductor current, the lower the overall ef-
ficiency of the converter. Unconvention-
ally, the BUZ10 is driven by the oscillator
test-output of the TL497A (pin 11) rather
than the internal output transistor.
Diode Di is another essential part in
the circuit. The requirements for this de-
vice are an ability to withstand high cur-
rent surges, and a low forward drop. The
Type BYV79 meets these conditions, and
must not be replaced with a general-pur-
pose type.
Returning to the circuit diagram of
Fig. 1, it should be borne in mind that
current peaks of 15-20 A are not uncom-
mon in the circuit. To prevent problems
arising with batteries having a relatively
high internal resistance, capacitor Cr
forms a buffer at the input of the conver-
ter. Since the converter charges the output
capacitors with short, surge-like current
pulses, two capacitors are connected in
parallel to ensure that stray capacitance
remains as low' as possible.
The power converter is not short-cir-
cuit resistant. Short-circuiting the output
terminals is the same as short-circuiting
the battery via Di and Li. The self-induct-
ance of Li is not so high as to limit the
current for the time required by a fuse to
blow.
A home-made inductor
Inductor Li is wound from 33V5 turns of
enamelled copper wire. Figure 5 shows
the dimensions. Most manufacturers sup-
ply enamelled copper wire on an ABS reel.
r
Fig. 5. Suggested construction of the in-
ductor on an ABS reel.
which is suitable as the former for making
the inductor. Drill two 2 mm holes in the
lower rim to pass the inductor wires: one
hole beside the cylinder and the other at
the outside of the rim.
There is little point in using thick w'ire
to wind the inductor, because the skin-ef-
fect, i.e., the displacement of charge car-
riers towards the outside of the w'ire, must
be taken into account given the frequen-
cies used in the converter. To ensure a low
resistance at the required inductance, it is
recommended to use tw'o wires of 1 mm
diameter, or even three or four wires of
0.8 mm diameter in parallel. Three
NEW PRODUCTS
9” Monochrome Monitors
9" Monochrome Monitors with compo-
site video nad for ITL input are now av-
ailable with reverse polarity protection
for 12V DC input. The Monitor has
Green Phospher Tube and has resolu-
tion of 800 x 35 video Amp. Bandwidth
of 22 MHz.
M/s. Anitex Marketing & Engineering
Co. Pvt. Ltd. • 234, Jaygopal Industrial
Estate • 510, Bhavani Shankar X Road •
Dadar • Bombav-400 028.
Hardware Locks
Real Time Systems have developed
Hardware Lock which preents unau-
thorized copies of software. This has in-
stallation software. Once installed, the
installed files can be freely copied but
will not run without the device in the
parallel part. The software contains its
own loader which does the loading and
hicrarchial decision making structure to
give maximum protection to software.
Further .no two units of installation
software are same for added security.
There is no limit to the number of files
that can be installed with one device. In
addition to this there is a data file protec-
tion unit DFP-1 which protects the prog-
ram source code, letters, reports and
0.8 mm wires result in a total diameter
that is roughly the same as that of two
1 mm wires, but has the advantage of re-
sulting in a 20% larger effective surface.
The inductor is close-wound and may
be encapsulated in a suitable resin or pot-
ting compound to limit the sound level
(remember that the frequency of oper-
ation is within the audible range).
Construction and alignment
The printed-circuit board designed for the
DC-DC converter is shown in Fig. 4. A
number of constructional points require
attention.
Resistors R 2 and Rn run fairly hot and
must, therefore, be mounted at a few mil-
limeters above the board surface. The
peak current through these resistors can
be as high as 15 A. The power-FET also
runs hot, and requires a medium-size
heat-sink and the usual insulating materi-
al. The diode can do without cooling, al-
though it is conveniently bolted on to the
same heat-sink as the power-FET (do not
forget to insulate it electrically). During
normal operation, the inductor heats up.
Heavy-duty terminals and wires must
be used at the input and output of the
converter. The battery is protected by a
16 A delayed action fuse inserted in the
input supply line. Remember that the fuse
does not protect the converter!
The circuit is simple to align: adjust Pi
for the desired output voltage between 20
and 30 V. The output voltage may be
made lower, but not lower than the input
voltage, by using a smaller resistor in po-
sition R4. The maximum output current is
about 3 A.
other data bases. Bothe thse units oper-
ate with IBM PC-DOS.
Real Time Systems • Plot No. 8, 4th Main
Road Avenue • Dhandeeswarar Nagar •
Velachery • Madras-600 042.
elektor india december 1989 1 2.45
PERSONAL COMPUTER DECISIONS
by Linda Bishop*
In choosing a pc system, the key question is not so much which
processor platform is the best’, but rather which is the most
appropriate platform for you. It is not simply a choice of
speed either. Memory access and multitasking capability
must also be considered in a platform decision.
And then, of course, there's software. What type of applications
will you run? What operating system do you need?
In software, as in the platform decision, several criteria
should be explored: price, performance, applications and the future.
OS / 2 addresses all these issues.
OS / 2 allows multi-tasking, multi-user
operation, breaks the 640 K barrier of Dos
and supports the graphical user interface
of presentation manager. This will make
network communication easier, provide
bigger databases, more complete and sim-
ple applications, and allow computers to
do several things at the same time.
What makes OS/2 unique is that it is
the first full-fledged multi-tasking system
for the 80286 microprocessor that can
switch back and forth between protected
mode and real mode to run the new pro-
grams designed for OS / 2 as well as most
existing DOS programs.This will give dos
users a smooth upgrade path to OS / 2.
The built-in network support of OS / 2
allows multi-user operation: this facility of
having several programs running at the
same time is, of course, a most useful one.
Moreover, OS/2 permits distributed ap-
plications, that is, it allows the program in
your pc to work (communicate) with pro-
grams in other pcs.
OS/2 was written for the 80286 pro-
cessor, taking advantage of the special
protected mode feature. This feature is
also provided by the 80386. OS/2 was
not written to take advantage of any of the
new features of the 80386 and no perfor-
mance advantages are obtained by running
OS/2 applications on an 80386.
The 80386 is no faster than an 80286
when running 16-bit software at the same
clock speed. The primary reason for this is
that the 80286 executes more 16-bit in-
structions in fewer clock cycles than the
80386 or 80386SX. Out of 190 existing
* Linda Bishop is a product marketing
engineer for Advanced Micro Devices’ Per-
sonal Computer Products Division,
Austin, Texas. She received her BSEE from
the University of Michigan (Dearborn).
Prior to joining AMO, she worked for
Motorola.
16-bit instructions, the 80286 is faster on
74, the 80386 is faster on 50 and the two
devices are the same on 66 instructions. In
fact, the only way the 80386 is able to run
OS / 2 at all is by emulating the 80286.
The applications that are available
today as well as those currently being
developed will not take advantage of the
80386 until an OS / 386 specific version
of the operating system is available some
lime next year: OS / 386 general applica-
tions are planned to become available
sometime in 1991-92.
Once an 32-bit operating system is
available for the 80386, the device will
have an advantage over the 80286. But
there is no guarantee that 80386, and espe-
cially 80386SX, personal computers avail-
able now have the configuration to run
new 80386 32-bit software four years
from now. After all, the first 80286-based
pc sold several years ago at 6 MHz with
640 K of memory is hardly suitable for
running 16-bit OS/2 now. The same situ-
ation is likely to exist in four years' time
for today's 80386 pc as far as running 32-
bit 80386SX software is concerned.
What is important for the OS / 2 oper-
ating system then is not whether it is run
on an 80286 or an 80386, but rather the
speed of the processor. The bulk of the
processor's work is multi-tasking, that is,
the accomplishing of several things at the
same time by dividing the computer's time
into ‘time slices’ that last only a fraction
of a second. These time slices are handled
so fast that it appears as if programs are
run simultaneously. Since the processor is
actually carrying out ali the tasks at sepa-
rate intervals (time slices), the faster the
processor, the quicker the multiple tasks
will be completed. An adequately equip-
ped 80286 system running at least 12-16
MHz with vga (Video Graphics Array)
graphics forms a very cost effective OS/2
foundation .
High-speed system pricing
80286 vs 80386
286-20
386-20
Dificrcnce
Dell
$2,999
$4,099
37%
Zeos
$2,095
$2,995
43%
Northgate
$2,599
$3,699
42%
PC Brand
$2,379
$2,995
26%
Dataworld
$1,555
$1,995
28%
CompuAdd
$1,695
$2,295
35%
Paradox OS/2 Benchmark
0 286 - 16 (U 386SX -16 386 • 16
890189-12
Figure 2
Display Write 4.0 OS/2 Benchmark
890189-13
Figure 3
1 2.46 elektor irtdia december 1989
The 80286 system offers everything
for the needs of today's and tomorrow's
user. Fast 80286 (16, 20 and 25 MHz) sys-
tems available now have the 16 Mbyte
memory access capability and the protect
mode for multiple applications required of
OS/2.
The 80286 is one of the best-selling
processors on the market today and it is
widely available. Moreover, its price is at
an economical level for the system
designer.
Owing to its die size, packaging and
complex processing, the 80386 is more
expensive. Moreover, systems built
around this device require 32-bit peripher-
als: the design cost is, therefore, higher as
well.
This leads to significant price differ-
ences between identically configured
80286-based and 80386-based personal
computers. As shown in the table, an
80386-based system costs on average
35% more than an 80286-based system.
The 80286 and 80386SX pcs used in
the tests to arrive at the comparison bar
graphs in Fig. 1, 2 and 3 are Everex stf.p
models, while the 80386 is an IBM System
80. The 80386 pc uses page mode memo-
ry access for 0.8 average wait states with
80 ns drams. Both the 80286 and the
80386SX run zero wait state with 60 ns
drams. The performance of these pcs is
indicative of that of other pcs.
The benchmark in Fig. 1 is based on
the R:Basc database program. The source
database used is PC Magazine's Index for
Volume 4.0. First, a Grouping Select
Query (SQL) was performed, followed by
a category tally to count the number of
PICTURE-IN-PICTURE
MINIBOARD FROM SIEMENS
The SDA 9088 Picture Insertion Processor
from Siemens allows the picture-in-picturc
facility to be installed not only in digital
tv sets, but also in analogue ones. The
need for only two chips reduces time and
material requirements and increases relia-
bility. The SDA 9088, which is designed
in Siemens 1 Mb it dram technology, also
provides a much better picture quality than
previous designs.
The SDA 9088 permits the insertion of
a reduced-size picture into the main pic-
ture by using picture signals that may be
based on completely different standards
and synchronization principles. The com-
bination of frame memory, control, digital
signal processor and digital-to-analogue
converters on a single chip enables equip-
ment manufacturers to realize the picture-
in-picture function in tv sets and video
recorders on a high-performance and par-
ticularly cost-effective basis.
occurrences in a category. Next, a calcula-
tion loop was performed on the first 100
records. The results are shown in seconds.
The bar graphs show that the 80286 pc
outperformed the 80386 pc by 4%, while
the 80386SX was 24% slower.
The bar graphs in Fig. 2 are obtained
from running the Paradox database pro-
gram on the three computers. The source
database is again PC Magazine’s Index
Volume 4.0. First, a Grouping Select
Query was performed. Next, a report was
run with the output sent to a file on ram
disk. The query results were then sorted
and a conditional delete of the records in
the query results was performed. The
results are shown in seconds. As is seen,
the 80286 pc was 18% faster than the
80386SX.
The comparative tests illustrated in
Fig. 3 were based on the ibm word proces-
sor program Display Write 4.2. The
benchmark started with a 100 K, 40-pagc
document. A global search and replace
was performed, changing one frequently
used word for another. Next, the margins
were narrowed, forcing a complete text
rewrap. Lastly, the document was repagi-
nated. The results arc shown in seconds.
Again, the 80286 pc was faster than the
80386 pc by 4%, while the 80386SX was
8% slower than the 80286 pc.
Comparative tests are influenced both
by the processor and by the memory inter-
face. In the pc systems used, the memory
interfaces were relatively equal (0.8 wait
states on the 80386 and 0 wait state on the
80286 and 80386SX machines). Thus, the
performance difference measured between
the 80286 and 80386SX was caused
ELECTRONICS SCENE
Although the picture-in-picture func-
tion has been in existence for some years,
it has failed to become widely established
in domestic video equipment owing to its
high cost, incurred mainly by the expen-
sive but indispensable frame memory and
the peripherals required for the analogue-
to-digital converters. Through the use of
the most up-to-date semiconductor tech-
solely by the different processors with the
former performing faster than the latter.
The performance difference between
the 80286 and 80386 must take into
account the different memory interface
techniques. A 0.8 wait state system (as on
the 80386 pc) has about a 9% perfor-
mance degradation compared to a true
zero wait state system (as on the 80286
pc). Taking this into account, the 80286
and 80386 systems performed essentially
the same.
As OS/2 software becomes more pre-
valent, pc performance will become more
important. Performance is primarily a
function of processor clock speed and
memory interface in the pc. Clock speeds
of 16 MHz and beyond will be needed to
run multiple applications effectively. It
should be borne in mind that there is little
difference in performance between the
80286 and 80386 running at the same
clock speed on OS / 2.
In addition to performance, price will
also remain a major factor in personal
computer decisions and it was seen that
80286-based pcs remain substantially
cheaper titan 80386-based systems. The
80286 has, moreover, a lot of life left for
dos, as well as OS / 2, systems and will
continue the trend toward higher clock
speeds.
According to Dataquest, the 80286
will increase its current market share of
IBM and compatible pcs from 30% to 33%
by 1992 and become the entry-level pc,
replacing 8086/8088 based machines.
Following a stable path to OS / 2, the
80286 is the best platform for cost vs per-
formance.
nology, it has now been possible to inte-
grate all essential functions into a single
circuit. The primary function of the pip is
to reduce the picture produced by the sec-
ond picture signal and synchronize it with
the main picture.
Two formats are available for the
inserted picture: 1/9 and 1/16 the size of
the main picture. The insert may be dis-
played in any of the four corners. A posi-
tioner for each corner permits adjustment
to the particular set's geometry.
In contrast to previous designs, picture
reduction is effected not by omitting the
pixels that are not needed but by digital
filtering of the horizontal and vertical sig-
nals to ensure that all the information is
utilized.
The SDA 9088 handles all worldwide
tv standards: a detector performs automat-
ic transfer to the standard being received.
It is also able to supply standard-converted
picture signals at a line frequency of
32 kHz.
elektor India december 1989 1 2.47
SPEEDING UP THE COMPUTER
by Pete Chown
The architecture of the
computer
If you look at a modern micro, say, an
80386-based IBM compatible, you will dis-
cover that nearly all the memory band-
width is used up. If faster memory were
installed, it might be possible to increase
the speed of the processor by several
times, but that would be the limit for that
particular architecture.
In an earlier article 1 I mentioned one
way out of this dilemma: parallel process-
ing. There are, however, many
other ways of speeding up appar-
ently sequential processors so that
they can reach speeds of up to 600
MFlops (million floating point
operations per second). At present,
the Cray-3 represents the limit of
that approach as far as commercial
machines go. The Cray-2 is the
fastest one that has been commer-
cially released.
Cacheing
Cacheing is one of the simplest
techniques that can be used to
speed up a computer. Earlier, I
mentioned that faster memory
could allow the speed of most ma-
chines to be increased substantial-
ly. Unfortunately, fast memory
costs a disproportionate amount
more, and so manufacturers decid-
ed to use the fast memory only for
instructions that are currently
being executed. This means that
the cache is loaded with the pages
of main memory that are being
used (normally in the opposite
phase of the processor clock to
that on which the processor reads
the memory), and it is then avail-
able for use.
Using a cache has one other
advantage. Memory protection - so that
one process can not alter another's memo-
ry - is very hard to implement fast enough
for the processor's request to access a par-
ticular word to be checked in time. On a
large machine, only of the order of 100 ns
would be available. If a cache is used,
however, the system can verify that the
process is allowed to use a particular page
before it is ever loaded into the cache. A
major cause of the inefficient use of
caches is that each time the machine
switches context (that is, changes the pro-
cess it is executing) at least part of the
cache has to be reloaded.
Multiple processors
Because large machines are generally used
for time-sharing, it is quite acceptable for
them to incorporate several processors.
Generally, however, these share the same
bus, so that problems are not encountered
with lack of memory on one processor, or
problems with an t/o device controlled by
another processor. Caches are used to
avoid continual conflicts for memory.
This tends to be a not very efficient
technique, because in practice a large
number of conflicts for memory do occur.
The best-known machine to use this sys-
tem is the vax 8900. It has four processors
sharing a bus (each of which is the same
as the single processor used in the 8700).
Adding a fourth processor does, however,
add only about 15% of the performance
that the processor would generate on its
own. The reason for this is that conflicts
for memory mean that the processors are
standing idle for much of the time.
The reason that dec decided to use this
technique is probably that it allowed them
to keep the same architecture: a radical
redesign would have meant changing the
instruction set, and the major selling point
of the vax range is that programs for any
vax can be run on any other. The other
advantage of this system is right at the top
end of the computer market: the US Navy
have produced a supercomputer using 1 6
largely independent processors, giving
them the edge over single-processor equi-
valents.
Pipelining and vector
processing
Pipelining and vector processing
are other major ways in which
manufacturers speed up their com-
puters. They are, however, much
more complex to implement than
the other systems. The techniques
are similar: some computers imple-
ment pipelines but not vector pro-
cessors, but generally speaking the
reverse is not true.
In pipelining, the processor, in-
stead of starting on one instruction
and executing it to completion,
reads instructions continually.
Once it has completed reading an
instruction, the processor begins
fetching the instruction's operands.
At the same time, the next instruc-
tion will be read, the previous
instruction will be executed, and
the result of the instruction before
that will be written to memory or
registers.
In practice, things are not this
simple. A pipeline tends to be
longer than just indicated, because
the aim is to keep the processor-
memory interface busy for as much
of the time as possible. Since not
all instructions need their operands
fetched, there would be a tendency
for the interface to run out of information
to fetch or store.
Problems with pipelines tend to be
encountered with jumps. When the proces-
sor jumps, everything in the pipeline is
useless because it no longer wants to exe-
cute those instructions. It is not possible to
make the pipeline start taking instructions
from the destination of the jump, because
the jump might be conditional and the
condition would not have been evaluated.
Another problem is when store loca-
tions change after the pipeline has been
loaded. If one instruction uses the result of
B90161 • 11
Construction of the Cray machines
1 2.48 eloktor india december 1989
the previous one, the old value that was
present at that location in store would
already have been loaded. There is no
solution to this except the long one - with
each and every instruction it must be
checked that the operand being loaded is
not going to be stored by an instruction
already in the pipeline. This is particularly
difficult with indirection, because care
must be taken that the information about
where the operand is coming from is avail-
able in time. If it is not, the processor must
stop until it is, which leads to inefficiency.
As with caches, pipelines suffer when a
processor switches context. Whereas with
the cache some of it might be able to be
preserved, the entire pipeline must be dis-
carded since there is nowhere for it to be
put until the processor returns to that pro-
cess.
Vector processors take the idea of pipe-
lining a stage further. With large machines
providing a large variety of complex
mathematical operations, the execution of
an instruction is by far the longest step in
the pipeline. Consequently, the informa-
tion about where to find the operands is
passed out to a lot of arithmetic proces-
sors. This saves the main processor from
having to find out what the operands are,
or to execute the instruction.
The problems with this are obvious.
The difficulties with making sure that the
operands of an instruction have not been
modified since the instruction was loaded
become much worse. Because some
instructions complete faster than others,
there is a danger of instructions being exe-
cuted in the wrong order: tens of short
instructions could have been executed in
the time it takes for a complex floating
point function to be evaluated and one of
these short instructions might have wanted
to use the result of the long one.
Another problem is memory bandwidth
- the multiple processor problems are
obviously much worse. This has, however,
been almost completely solved. Memory,
instead of being addressed over a single
bus, is addressed on a chip-by-chip basis,
so that as long as all the processors wish to
access different chips, they can do so at
the same time. This solution does, howev-
er, lead to another snag: the large amount
of wire needed to connect each individual
chip!
It is interesting to note that this archi-
tecture is based on parallel processing,
even though the machines appear sequen-
tial to the user. The parallelism is on a
very small scale, and so it has been
described as ‘fine’ parallelism, whereas
true parallel processing machines have
been described as having ‘coarse’ paral-
lelism.
As these computers get faster, the exact
length of wire used to connect two points
becomes significant in determining timing.
Consequently, Cray Research decided to
cut each piece of wire in their machines
the same length! Unfortunately, these
lengths have to be also as short as possible
for the same reason and this led to the cir-
cular construction of the Cray machines as
illustrated. It also led to the situation
where the wires are almost impossible to
get at, forming a three-dimensional web of
cables that are tight enough for it to be dif-
ficult to reach a wire near the middle.
RISC processors
Rise processors are not really viable as a
technique for building large machines.
The reason is that you are faced with a
choice of ways of improving performance
- make each instruction do more or exe-
cute faster. Small machines had been tend-
ing to follow the former route despite the
fact that there was not really enough pro-
cessing power on a single chip to do it. A
large increase in speed was therefore ob-
tained when micros began to follow the
latter route. Large machines have
pipelines, caches and so on, and also aim
to do a lot per instruction. Consequently,
the Sun, Apollo and Hewlett Packard
machines tend to set the limit for this type
of technology.
There is now a move to provide a
mainframe style processor on a chip, since
this is becoming viable with greater relia-
bility and packing density. This will effec-
tively make the Rise processor obsolete in
a few years’ time, at least as far as the very
fastest workstations are concerned.
This trend towards micros that are
more like mainframe is actually another
way of speeding up computers. We are ap-
proaching the limit as far as supercomput-
ers go. but if workstations that only sever-
al people use get nearly that fast, they will
effectively have a much more powerful
machine because there are far fewer pro-
cesses for it to run.
There will always be a place for the
supercomputer, however, in performing
single processes that are too complex for a
workstation to do. It will, however, be-
come increasingly wasteful to use a super-
computer for a lot of fairly small jobs.
One area of potential for Rise that has
not received much attentional is that of
arithmetic processing. It would be possible
to build a Rise machine with, say, 256
bytes of ram and several registers that
would carry out operations between regis-
ters only and not ram. It would thus be
very simple and could, therefore, run at
high speeds. It could then be programmed
with short, repetitive calculations that
could be done over and over again.
Managing a pipeline
I have already discussed some of the prob-
lems that arise from pipelining and vector
processing. One of the easiest ways to
understand the problems and how they are
solved is, however, to look at how a vector
processor would execute a certain
sequence of instructions.
Since this is only for illustration, the
instructions will be given in words - not in
any form of mnemonic that would make it
harder to follow. The instructions are to
calculate the coordinates needed to draw a
circle by trigonometry. Square brackets
indicate indiscretion The label ‘pointer’
points to a location containing the address
where the forty pairs of coordinates are to
be placed.
1. Load register A with 0.
2. Load register B with 0.
3. Label:
4. Calculate cos(A), pul in register C.
5. Calculate sin(A), put in register D.
6. Multiply C by [radius).
7. Multiply D by [radius].
8. Store register C at [pointer] + B.
9. Store register D at [pointer] + B + 1 .
10. Add 2 to B.
1 1 . Add pi/20 to A.
12. Jump to label if A < 2 * pi.
Let us now' consider how a vector pro-
cessor would execute this section of code.
It would start by filling its pipeline from
the beginning. No evaluation of operands
would be necessary for instructions 1 and
2. When these got to be executed, they
will be run at the same time because the
processor would recognize that (hey did
not refer to the same part of store.
Instructions 4 and 5 could not be exe-
cuted until instructions 1 and 2 had been
completed, because the values of the same
registers are used. Once 1 and 2 had been
completed, however, they would be exe-
cuted together.
The same would be (rue of instructions
6 and 7, but here one of the advantages of
a fast processor shows up. The processor
has been instructed to look at a particular
memory location in order to find the
radius of the circle. There is no reason
why this should wait to be evaluated until
the rest of the instruction can be. Different
processors would tackle it in different
ways: those with just a pipeline and no
vector processor would attempt to find
time to evaluate it while the instruction is
in the pipeline, while those with a vector
processor would simply hand the pointer
to one of the arithmetic units and instruct
it to look at that place in store.
The two additions would take place
concurrently, since they do not refer to
each other in any way. The jump would
then be encountered. The pipeline would
have been unable to follow the jump to its
conclusion to get subsequent instructions,
because it is a conditional one. It is, there-
elektor India decamber 1989 1 2.49
assume that the jump will not be taken,
and it will have to abandon all the infor-
mation it has built up about the instruc-
tions following the loop, except when the
loop finally ends. Nothing has been lost
compared to a conventional processor,
however, because the bus would merely
have been sitting idle. Once back at the
start of the loop it might have kept the
instructions because such an eventuality
was likely or it might have to start build-
ing up its pipeline from scratch again.
Conclusions
Because we are reaching the limits of
semiconductor-based computers, the large
computer of today is a far more complex
thing than its predecessors. The normal
rules of structured design have been aban-
doned in a search for the last megaflop,
leading to such peculiarities as computers
with all the wires the same length (nor-
mally, of course, no one would think of
building a large system other than in stan-
dard 19 in. rack-mounted cases on a care-
fully constructed backplane). The tech-
niques do, however, work and we have
probably got computers an order of mag-
nitude faster from them. It is, however, a
tribute to the people who design them that
they work at all.
■HMMMj
NfclW PRODUCTS
Hand Cleanser
Advance Labs have introduced Actoplus
Hand Cleanser. This remove grease, oil,
smal particles of metal, dust, grime and
dirt instantly when applied. A small
quantity is applied in paste form and
either washed away wth water or simply
wiped clean with cotton waste or cloth.
There are no side effects as it is abso-
lutely safe.
Advance Lab • 11, Below Shantidoot
Hotel • Dr. Ambedkar Road • Dadar •
Bombay- 400 014.
Sequential Timers (Cyclic)
Vectrol Engineers introduces a solid
state cyclic timers. These are available
with two change over contacts. Timer
switches a given device/load sequentially
“ON-OFF” on giving signal/application
of control supply and stops it when a
soap signal/command is given. Timers
can be supplied with ‘ON’ time count or
‘OFF’ time count start first on applica-
tion of start command/signal/control
supply to timer. Timer provides precise
‘ON-OFF’ sequence ratio with excellent
repeat accuracy.
Cyclic timers are used in chemical/phar-
maceutical and other allied industries,
where device/load is required to repeat
the operation automatically in succes-
sion, until the stop signal is given.
Vectrol Engineers • 4 A/32, Versova
View 1 Co-op. Hsg. Society • Four-
Bunglow road • Andheri (W) • Bombay-
400 058.
Grasslin Time Switch
The MIL 2008 Q series is fitted with a
Quartz Electronic Drive Control and a
step motor. The Quartz frequency is 14.9
million Hertz and the Quartz stabiliza-
tion guarantees the exact running of the
driving mechanism. These time switches
are designed for the accurate and effort-
less control of oil heating installations,
electric heaters, air conditioning plants,
water processing plants, street lights,
traffic signals, etc., etc.
Human nature being what it is, howev-
er. these techniques will probably be with
us even when optical computers appear,
and we will simply take our thousand
times speed increase, and do exactly the
same with optical fibres.
MIL 2008 Q is available with contact rat-
ing of 16Amps. 250V AC and available
with daily programme and weekly prog-
ramme dial. Operates on mains supply
and continue to run for 150 hrs. after
power failure on a battery back-up.
M/s. Sai Electronics • (In association
with Cupwud Arts) • Thakore Estate •
Kurla Kirol Road • Vidyavihar (West) •
Bombay-400 086. Ph: 5136601/5113094/
5113095.
12.50 elektor india december 1989
( 2 )
The first part of this article dealt with the design
considerations concerning loudspeaker crossover
filters in general, and active crossover filters in
particular. This month a practical circuit is given,
with details on how to modify it according to
personal taste.
As explained last month, several de-
cisions must be made before starting
with the actual design of any loud-
speaker crossover filter system. In
chronological order:
— What type of filters: active only,
hybrid or passive? This article only
deals with filters that are active, at
least in part.
— What type of system, three-way or
two way? This decision will be based
on such factors as desired cabinet
size, available financial resources,
desired frequency range — and
personal taste.
- Which speakers? This depends in
part on the answer to the previous
question.
- What crossover frequencies, and how
steep the filters? These decisions are
both based on the answer to the
previous question.
- Which amplifiers? This is a source of
endless discussion, but the answer
obviously depends in part on the
type of system and the speakers used.
The points of interest in this article are
the design decisions for the filter proper:
two-way or three-way, what crossover
frequency or frequencies, and how
steep? These points are illustrated in
figure If. If a two-way system is
required, the crossover frequency is
assumed to be f 1 — f2 can be ignored.
For a three-way system, fl is the lower
crossover frequency and f2 is the higher.
The filter slopes ■ can be 6-, 1 2- or
18 dB / octave, and the 12- and
18 dB/octave slopes are numbered in
figure If.
As an example, a three-way system with
crossover frequencies of 400 Hz and
4 kHz and filter slopes of 12 dB/octave
at the lower crossover point and
18 dB / octave at the higher fre-
quency can now be defined briefly as
‘fl = 400 Hz, f2 = 4 kHz, filter slopes
1, 4, 6 and 7’. This shorthand notation
will be used extensively in the tables
given in this article.
The most complex circuit diagram is
given in figure 5: a three-way system
with all slopes 18 dB/octave. This
corresponds to the figure 6 layouts for
printed circuit board and parts.
When any less-complex set up is to be
assembled it will only be necessary to
complete the ‘through paths’ with wire
links on the printed circuit board. This
will be illustrated in detail further on.
For added convenience, all the circuits
and parts-layouts have been duplicated
several times — each time showing the
simplified schemes and jumper wires
needed for the less complex filters. The
schemes we have chosen to illustrate
are:
— Three-way system with 1 2 dB/octave
slopes (figures 7 & 8).
— Two-way system with 18 dB/octave
slopes (figures 9 & 10).
— Two-way system with 12 dB/octave
slopes (figures 1 1 & 1 2).
— Two-way system with 6 dB/octave
slopes (figures 13 & 14).
The frequency responses of the figure 5
filter set are plotted in figure 15.
Figure 16 gives the plots for the figure 7
circuit. In both cases the frequencies
chosen for illustration are 500 Hz (fl)
and 5 kHz (f2).
Design procedure
The suggested procedure for finding the
required design is as follows. First of all
decide, using figure 1 f or table 1 , which
set of filter characteristics is to be
realised - and which crossover fre-
quences ( values of fl and f2) are to be
taken. Table 2 may now be used as a
kind of ‘railway timetable’ to determine
which PC board positions are to be left
■open, which positions must be bridged
by a jumper wire and which of the
tables 3 ... 8 is to be referred to for the
component values. The examples given
will illustrate this.
Loudspeaker connections
In just the same way as with passive
filters it is important to connect the
individual loudspeakers in the correct
relative phases. The rules are as follows:
When the filter provides a three-
way symmetrical crossover with
12 dB/octave slopes, the midrange
unit should be connected in opposite
sense to the woofer and tweeter.
Both systems of a stereo pair should
of course be identically wired.
®
©
©
L
M
H
©
©
©
Figure If. A few frequency-response plots,
with slopes of 12 and 18 dB/octave and one
or two crossovers, as an aid to interpretation
of table 1.
If
elektor india december 1989 12.51
- When the filter provides a symmetri-
cal two-way crossover with
12 dB/octave slopes, the tweeter
should be connected in opposite
sense to the woofer-midrange unit.
© 0
L H
0 ©
— The problem is different with
18 dB/octave and 6 dB/octave slopes,
where the phase shift in the filters at
crossover totals 270° or 90°. It is
convenient to connect all speakers in
the same sense in these cases.
The loudspeaker-coupling electrolytic
capacitors in the midrange and treble
channels can in principle be given a
smaller value than that in the woofer
channel, thus saving space and cost.
However, one must bear in mind that a
smaller value component will have a
lower alternating current (‘ripple’)
rating. The smallest value that still has
a current rating at least equal to the
loudspeaker maximum RMS current
will usually have a large enough
capacitance too. In case of doubt ensure
that the RC cutoff point of the
12.52 elektor india december 1989
capacitor with the loudspeaker’s
nominal impedance is 3 ... 5 times
lower than the high-pass crossover
frequency in the channel concerned.
The factor 3 ... 5 should also be
observed with the woofer! This results
in the well-known rule of thumb:
where f c is the lower crossover fre-
quency.
Nothing useful is gained (and there is a
risk of too much phase shift or
amplitude rolloff being caused) by also
reducing the values of the input coupling
capacitors of the midrange and treble
amplifiers. Cl 6 and C21 in the filter are
‘unnecessarily large’ for the same reason.
One final remark concerns the function
of the presets PI , P2 and P3. These are
not intended as tone control adjust-
ments! They should be used only to
compensate for possibly unequal sensi-
tivities of the individual amplifier-
speaker channels. Deliberate maladjust-
ments of not more than 3 dB (tone
controls after all!) may however
occasionally be permissible.
Component list for figures 5 and 6.
Resistors:
R1 f R2
= 220 k
R3.R8.R14.
R19 l ,R24‘
= 5k6
R4.R9.R15,
R20 1 ,R25 !
= 2k2
R5 2
see table 3
R6 3
see table 3 or 5
R7
see table 3, 5 or 7
RIO 4
see table 4
R1 1 5
see table 4 or 6
R12 R13
see table 4, 6 or 8
R16 5,6
see table 3
R17 3,6 ,R18'
see table 3 or 5
R21 1,4
see table 4
R22 1 ,R23 1 ,R26 I
see table 4 or 6
P1.P2.P3'
10 k preset
Capacitors:
Cl
= 470 n
C2.C6.C1 1 .
C15 1 .C20 1
= 4n7
C3 4
see table 3
C4 5
see table 3 or 5
C5
see table 3, 5 or 7
C7.C16.C21 1
= 10 p/25 V
C8 1
see table 4
C9 3
see table 4 or 6
CIO
see table 4, 6 or 8
C12 1 ' 4
see table 3
C13 6 .C14'
see table 3 or 5
Cl 7°
see table 4
C18 1 .C19 1
see table 4 or 6
C22
= 100 p/40 V
C23.C24.C25,
C26\C27'
= 100 n
Semiconductors:
T1.T3.T5.T7 1 ,
BC107 B, BC547 B
T9 1
or equivalent
T2.T4.T6.T8 1 ,
BC177 B, BC557 B
T10 1
or equivalent
Footnotes
means: omit part for two-way filter
set
means: replace by wire link for
12 dB/oct and 6 dB/oct.
3 means: replace by wire link for
6 dB/oct.
4 means: omit this part for
12 dB/oct or 6 dB/oct.
5 means: omit this part for
6 dB/oct.
6 means: replace by wire link for
two-way filter set.
NB. The 6 dB/octave slopes are only
useful in a very limited number of
two-way system designs— the tables
therefore do not give values for three-
way design.
Figure 15. Frequency response of the figure 5
circuit, as measured with fl set at 500 Hz and
12 at 5 kHz.
Figure 16. Frequency response of the figure 7
circuit with the same crossover points as
figure 15.
Figure 5. Complete circuit diagram of an
active filter set for two symmetrical
18 dB/octave crossovers (three-way).
Figure 6. Component layout and p.c. board
copper-side plan for the figure 5 circuit.
(EPS 9786)
eloktor india december 1989 1 2.53
Table 3.
Table 1.
The 18 dB/octave low-pass filter, having the response given in
figure 2a
, with
the nominal
crossover
frequencies obtainable
The different possible combinations of symmetrical
or asym-
using El 2 series component values.
metrical crossovers and 12 or 18 dB/octave slopes.
f (Hz)
R (k£i)
c a (nF)
C b (nF)
C c (nF)
filters slopes at
filters slopes at
fi
R5
R6
R7
C3
C4
C5
fi to be
be
combine from
refer to
f2
R16
R17
R18
Cl 2
C13
C14
V*
figure If
figures
97
10
10
10
220
560
33
18 12
18
18
2, 4, 6 & 7
119
10
10
10
180
470
27
18 12
12
12
2,4.5 8< 8
146
10
10
10
150
390
22
18 12
18
12
2, 4, 6 8i8
179
10
10
10
120
330
18
18 12
12
18
2,4, 5& 7
214
10
10
10
100
270
15
12 18
18
18
1,3, 6& 7
268
10
10
10
82
220
12
12 18
12
12
1,3, 5 8i8
322
10
10
10
68
180
10
12 18
18
12
1,3, 6& 8
392
10
10
10
56
150
8.2
12 18
12
18
1 , 3, 5 8t 7
472
10
10
10
47
120
6.8
18 18
18
18
2, 3, 6 8i 7
5 & 6
574
10
10
10
39
100
5.6
18 18
12
12
2, 3, 5 Si 8
684
10
10
10
33
82
4.7
18 18
18
12
2.3, 6& 8
824
10
10
10
27
68
3.9
18 18
12
18
2, 3, 5 & 7
974
10
10
10
22
56
3.3
12 12
18
18
1 , 4, 6 & 7
1191
10
10
10
18
47
2.7
12 12
12
12
1,4, 5& 8
7& 8
1461
10
10
10
15
39
2.2
12 12
18
12
1,4,6 & 8
1786
10
10
10
12
33
1.8
12 12
12
18
1,4, 5 & 7
2143
10
10
10
10
27
1.5
18 18
—
2&3
9& 10
2679
10
10
10
8.2
22
1.2
12 12
—
1 & 4
11 & 12
3215
10
10
10
6.8
18
1
12 18
—
1 8i 3
3921
8.2
8.2
8.2
6.8
18
1
18 12
-
2 & 4
4728
6.8
6.8
6.8
6.8
18
1
5742
5.6
5.6
5.6
6.8
18
1
6841
4.7
4.7
4.7
6.8
18
1
8244
3.9
3.9
3.9
6.8
18
1
9743
3.3
3.3
3.3
6.8
18
1
response
(see figure If)
1
2
3
4
5
6
7
8
9
10
component
-V
R5
t3
wl
wl
R6
t3
t5
wl
R7
t3
t5
t7
C3
t3
—
—
C4
t3
t5
—
C5
t3
t5
t7
C8
wl
t4
wl
C9
t6
t4
wl
CIO
t6
t4
t8
R10
—
t4
—
R11
t6
t4
—
R12
t6
t4
t8
R13
t6
t4
t8
R16
t3
wl
R17
t3
t5
R18
t3
t5
C12
t3
—
C13
t3
t5
C14
t3
t5
Cl 7
wl
t4
C18
t6
t4
Cl 9
t6
t4
R21
—
t4
R22
t6
t4
R23
t6
t4
R26
t6
t4
see figure
3b
2b
2a
3a
3b
2b
2a
3a
4a
4b
Cross-reference table of frequency-determining components,
starting from the 'available response curves' of figure If. The
components are numbered as in the complete circuit and layout
diagrams (figures 5 & 6); t3 t8 are the value-table references,
'wl' means 'wire link' and ’ means ’omit'.
The 18 dB/octave high-pass filter, having the response given in
figure 2b, with the nominal crossover frequencies obtainable
using El 2 series component values.
C {nF}
C8 = C9 = CIO
C17 = C18 = C19
100
82
68
56
47
39
33
27
22
18
15
12
10
8.2
6.8
5.6
4.7
3.9
3.3
2.7
2.2
1.8
1.5
1.2
1
f (Hz)
R a (kn)
Rb (kn)
R c (kn)
fi
R10
R11
R1 2 = R13
f2
R21
R22
R23 = R26
114
10
3.9
150
139
10
3.9
150
168
10
3.9
150
204
10
3.9
150
243
10
3.9
150
293
10
3.9
150
346
10
3.9
150
423
10
3.9
150
519
10
3.9
150
635
10
3.9
150
762
10
3.9
150
952
10
3.9
150
1140
10
3.9
150
1390
10
3.9
150
1680
10
3.9
150
2040
10
3.9
150
2430
10
3.9
150
2930
10
3.9
150
3460
10
3.9
150
4230
10
3.9
150
5190
10
3.9
150
6350
10
3.9
150
7620
10
3.9
150
9520
10
3.9
150
11400
10
3.9
150
12.54 elBktor India december 1989
How to use the tables.
• Decide on the type of filter required,
and refer to figure If and/or table 1
for the ‘shorthand notation’. Note that
responses 9 and 10 are 6 dB/oct low-
pass and high-pass, respectively; these
are not shown in figure 1 f.
• Proceed to table 2. Under each of the
(two or four) chosen response curves,
further information is given regarding
a group of frequency-determining
components. This can be either *wl’
(wire link), ' (omit) or reference to
one of the tables 3 ... 8 (e.g. ‘t3’
means ‘refer to table 3’).
• Proceed to the tables referred to. As
an example, assume that slope 3 is
required at a lower crossover frequency
f 1 = 400 Hz. Under response 3, table 2
refers to table 3 for R5 . . . R7 and
C3 . . . C5. Proceeding to table 3, the
nearest frequency to the desired
400 Hz is 392 Hz. For this frequency,
the values of R5 . . . R7 are shown as
10 k«, C3 = 56 n, C4 = 150 n and
C5 = 8n2.
Bibliography
Electronics, August 18th 1969, p82 etc
(filter circuits)
J.R. Ashley & L.M. Henne:
Operational Amplifier Implementation
of Ideal Electronic Crossover Networks;
JAES, January 1971.
S. Linkwitz: Active Crossover Networks
for Noncoincident Drivers;
JAES, February 1976.
J.R. Ashley & A.L. Kaminsky:
Active and Passive Filters as Loud-
speaker Crossover Networks; JAES,
June 1971.
R.H. Small: Constant-Voltage Crossover
Network Design; JAES, January 1971.
B.B. Bauer; Audibility of phase
distortion; Wireless World, March 1974.
H.D. Harwood: Audibility of phase
effects in loudspeakers; Wireless World,
January 1976.
Table 5.
The 12 dB/octave low-pass filter, having the
response
given in
figure 3a,
with the
nominal
crossover
frequencies
obtainable
using El 2 series component values.
f (Hz)
R (kn)
C b (nF)
C c (nF)
fl
R6 = R7
C4
C5
f2
R1 7 = R1 8 C13
C14
102
22
100
47
125
18
100
47
150
15
100
47
188
12
100
47
225
10
100
47
274
10
82
39
331
10
68
33
402
10
56
27
479
10
47
22
577
39
10
4.7
682
33
10
4.7
834
27
10
4.7
1020
22
10
4.7
1250
18
10
4.7
1500
15
10
4.7
1880
12
10
4.7
2250
10
10
4.7
2740
10
8.2
3.9
3310
10
6.8
3.3
4020
10
5.6
2.7
4790
10
4.7
2.2
5840
8.2
4.7
2.2
7040
6.8
4.7
2.2
8550
5.6
4.7
2.2
10190
4.7
4.7
2.2
Table 7.
The 6 dB/octave low-pass filter, having the
response
given in figure 4a,
with the
nominal
crossover frequencies
obtainable
using E 1 2 series component values.
f (Hz)
R (kfl)
C c (nF)
fl
R7
C5
106
10
150
133
10
120
159
10
100
194
10
82
234
10
68
284
10
56
339
10
47
408
10
39
482
10
33
589
10
27
723
10
22
884
10
18
1060
10
15
1330
10
12
1590
10
10
1940
10
8.2
2340
10
6.8
2840
10
5.6
3390
10
4.7
4080
10
3.9
4820
10
3.3
5890
10
2.7
7230
10
2.2
8840
10
1.8
10600
10
1.5
Table 6.
The 12 dB/octave high-pass filter, having the
response
given in
figure 3b
with the
nominal
crossover
frequencies
obtainable
using El 2 series component values.
f (Hz)
C (nF)
Rb (kn)
R c (k£2)
fl
C9 = CIO
R 1 1
R1 2 = R13
f2
C18 = C19
R22
R23 = R26
113
100
10
39
137
82
10
39
165
68
10
39
201
56
10
39
239
47
10
39
289
39
10
39
341
33
10
39
417
27
10
39
511
22
10
39
625
18
10
39
750
15
10
39
938
12
10
39
1130
10
10
39
1370
8.2
10
39
1650
6.8
10
39
2010
5.6
10
39
2390
4.7
10
39
2890
3.9
10
39
3410
3.3
10
39
4170
2.7
10
39
5110
2.2
10
39
6250
1.8
10
39
7500
1.5
10
39
9380
1.2
10
39
11300
1
10
39
Table 8.
The 6 dB/octave high-pass filter, having the
response
given in figure 4b,
with the
nominal
crossover frequencies
obtainable
using El 2 series component values.
f (Hz)
R c <k«)
C(nF)
fl
R1 2 = R13
C19
106
22
150
133
22
120
159
22
100
194
22
82
234
22
68
284
22
56
339
22
47
408
22
39
482
22
33
589
22
27
723
22
22
884
22
18
1060
22
15
1330
22
12
1590
22
10
1940
22
8 2
2340
22
6.8
2840
22
5.6
3390
22
4.7
4080
22
3.9
4820
22
3.3
5890
22
2.7
7230
22
2.2
8840
22
1.8
10600
22
1.5
elekior india december 1989 1 2.55
<HH> oG T
I
o^o
f o
5<HH>
<L^> 2o-^o6geHH»f|
i -i °£”3j°
i r^-x < H rib i *o < H^g3^ 6
V < ^^19 h o o 1we.a| 0
<£V °HH>S <Hl-oS
njOHI-O
<Hho 2
<Hhog
■BHEKj
^<>-Qh.-P $
©‘Sfc 0 s-i
o oHho o- — {] | — o
-*• • •.. cs6 t
©15 ...30V
T1,T3,T5,T7,T9 = BC547B, BC107B
T2.T4.T6.T8.T10 = BC557B. BC177B
3 - way, 12 dB/oct.
As an example, assume that a three-way
12 dB/oct. filter system is required
(slopes 1, 4, 5 and 8 in figure If) with
crossover frequencies fl = 400 Hz and
f2 = 3 kHz.
Referring to table 2: for slope 1,
C8 = wire link; R1 0 = omitted ; C9, Cl 0,
R 1 1 ... R 1 3 are to be found from
table 6. In the latter table, the nearest
frequency to the desired fl is 417 Hz.
The corresponding component values
are given as C9 = CIO = 27 n;
kll = 10 k; R12 = R13 = 39 k.
Back to table 2: for slope 4, R5 = wire
link; C3 = omitted; R6, R7, C4 and C5
are to be found from table 5. Proceeding
to this table, the component values
corresponding to f 1 = 402 Hz are shown
as R6 = R7 = 10 k; C4 = 56 n and
C5 = 27 n.
Back to table 2 : for slope 5 , C 1 7 = wire
link; R21 = omitted; C18, C19, R22,
R23 and R26 are to be found from
table 6. For f2 = 2890 Hz (the closest to
the desired 3 kHz), this table gives the
component values: C18 = C19 = 3n9;
R22 = 10 k; R23 = R26 = 39 k.
Now table 2 again: for slope 8,
R16 = wire link; C12 = omitted; R17,
R18, C13 and C14 are to be found from
table 5. For f2 = 2740 Hz, this results in
R17 = R18 = 10k; C13 = 8n2;
Figure 7. Circuit diagram of an active three-
way filter with symmetrical 12 dB/octave
crossovers.
Figure 8. Parts layout modified for the
figure 7 circuit.
C14 = 3n9.
Finally, referring to the parts list for
figure 6 gives all other component
values. Note that the footnotes 2 and 4
are valid in this case (12 dB/oct);
however, we had already found these
wire links and omitted parts from
table 2.
12.56 elektor india december 1989
oH |-o o^fTSj
^ h , ■« £ .. •
ra<H (-O
goHho
j <HH>8
joHH>8
} <H h° £ °
T1.T3.T5, - BC547B, BC107B
T2.T4.T6. = BC557B, BC177B
15. 30V
B...15mA
2-way, 18 dB/oct.
The two-way filter is assembled on the
same board. In this case T6 collector has
to be linked with the ‘hot’ side of C 16-
no matter which filter slopes are chosen
— and the gain of the ‘high’ channel is
preset by P2.
Correct use of the tables should produce
this result automatically. As an example,
assume that slopes 2 and 3 are required
at a crossover frequency fl = 500 Hz.
For slope 2, table 2 refers to table 4 for
the following components: C8 ... CIO
and RIO . . . R13. For slope 3, the table
refers to table 3 for R5 . . . R7 and
C3 . . . C5.
Proceeding first to table 4, the
component values for fl = 519 Hz are
found to be: RIO = 10 k, Rll = 3k9,
R12=R1 3= 150k,C8 = C9 = C10 = 22n.
Referring now to table 3, the com-
ponent values for fl = 472 Hz are found
to be: R5 = R6 = R7 = 10 k; C3 = 47 n;
C4 = 120 n; C5 = 6n8.
Finally, the parts list for figure 6 gives
all other components. Footnote 1 is
valid in this case: ‘omit this part for
two-way filter set’. This turns out to
mean that T9 and T10 (figure 5) are
omitted, with all associated components;
T7 and T8 are also omitted, with all
associated components. Furthermore,
footnote 6 is valid: ‘replace by wire link
Figure 9. Two-way circuit with symmetrical
18 dB/octave crossover.
Figure 10. Parts layout modified for the
figure 9 circuit.
for two-way filter set’. This refers to
R16, R17 and C13, giving the required
through path from T6 to Cl 6. Note
however that on the component layout
a single wire link is shown, direct from
one end of R16 to one end of C13. This
will also work of course . . .
elektor India december 1989 1 2.57
2-way, 12 dB/oct.
In figure If, the required slopes are
numbered 1 and 4. Assume that the
crossover frequency is to be fl = 1 kHz.
As before, the first table to look at is
table 2. For slopes 1 and 4, C8 and R5
both have to be replaced by wire links;
RIO and C3 are omitted; the values for
C9, CIO and Rll . . . R13 are to be
found from table 6; the values for R6,
R7, C4 and C5 are to be found from
table 5.
First table 6. For fl = 938 Hz, the
component values are given as follows:
C9 = CIO = 12 n; Rll = 10 k;
R12 = R13 = 39 k.
Now table 5. Here the nearest frequency
given is f 1 = 1 020 Hz. The corresponding
‘ 22 k;
component values are: R6 = R7 ;
C4= 10 n;C5 = 4n7.
Finally, check the parts list. In this case,
footnotes I, 2, 4 and 6 are all valid. In
other words, components marked either
1 or 4 are to be omitted and components
marked either 2 or 6 are to be replaced
by wire links.
To sum it up, the complete parts list for
this example would be:
Resistors:
R1.R2 = 220 k
R3.R8.R14 = 5k6
R4.R9.R1 5 = 2k2
R5 - wire link
R6.R7 = 22 k
Rll = 10 k
R12.R13 = 39 k
R16.R17 = wire link
PI ,P2 = 10 k preset
Capacitors:
Cl = 470 n
C2.C6.C11 = 4n7
C4 = 10 n
C5 = 4n7
C7.C16 = IOp/25 V
C8 = wire link
C9.C10 = 12 n
C13 = wire link
C22 = 100 p/40 V
C23.C24.C25 = 100 n
Semiconductors:
T1.T3.T5 = BC107 B or equivalent
T2.T4.T6 = BC177 B or equivalent
I 2.58 elektor india december 1989
Figure 11. Two-way circuit with symmetrical
12 dB/octave crossover.
Figure 12. Parts layout modified for the
figure 1 1 circuit.
Figure 13. 6 dB/octave two-way circuit dia-
gram.
Figure 14. Parts layout modified for the
figure 13 circuit.
2-way, 6 dB/oct.
Before going any further, it should be
stated clearly that 6 dB/oct slopes are
only useful in a very limited number of
applications. They should be used with
caution, since there is always a danger
of destroying the high-range loud-
speaker.
However, for completeness’ sake an
example is given here: two-way, 6 dB/oct
(slopes 9 and 1 0, not shown in
figure If), with a crossover frequency
fl = 4 kHz.
Table 2 specifies a wire link for R5, R6,
C8 and C9; C3, C4, RIO and R1 1 are to
be omitted. The values for R7 and C5
are to be taken from table 7; the values
for CIO, R12 and R13 are to be taken
from table 8.
For fl = 4080 Hz, table 7 specifies
R7= 10 k and C5 = 3n9.
For fl = 4080 Hz, table 8 specifies
R12 f R13 = 22 k and C19 = 3n9.
In this case, all 6 footnotes in the parts
list are valid . . . Since footnotes 1, 4
and 5 are valid, the following com-
ponents should be omitted: R10, Rll,
R18 . . . R26; P3; C3, C4, C12,
C14, C15, C17...C21, C26, C27;
T7 . . . T10. Furthermore, since foot-
notes 2, 3 and 6 are valid, the following
components are to be replaced by wire
links: R5, R6, R16, R1 7; C8, C9, C13.
Note that Cl 7 has already been
eliminated by footnote 1, and is there-
fore not replaced by a wire link when
we get to footnote 2! H
Hlektor India december 1989 1 2.59
TRAVELLING-WAVE TUBES
B. Higgins
Although many electronics engineers are not familiar with their basic
operation and applications, travelling-wave tubes (TWTs) are
important components used in satellites and other microwave
applications. Their use has increased rapidly in line with the
widening of the available radio spectrum and the continuing
development of satellite communications systems. Recently
commissioned medium and high-power TV satellites such as Astra
1 A, DFS Kopernikus, TV-SAT2, TDF-1 all use high-performance TWTs
to provide television pictures around the clock to millions of viewers.
A travelling-wave tube is an electronic
amplifier for microwave radio signals. It
is not, strictly speaking, a thermionic tube,
but rather a complete wideband RF power
amplifier in a vacuum envelope. Origin-
ally developed in the mid 1940s, TWTs
have been improved considerably since
then. In particular, their power efficiency
has gone up over the years from a modest
1 0 to 20% to nearly 50% for the latest types
used in direct-broadcasting TV satellites.
The radio signals produced by TWTs
are normally in the frequency range from
2 GHz to 22 GHz, spanning the S, C, X, Ku
and Ka bands. Table 1 lists the 10 different
TWTs operating at frequencies spread
across these radio bands.
The outstanding feature of the TWT is
its high power gain of 30 dB to 55 dB. This
means that an input power of less than
1 mW is sufficient to achieve an output
power of tens of watts across a wide fre-
quency range. Disadvantages of the TWT
are its size and weight, relatively low effi-
ciency, and high-voltage power supply re-
quirement.
How it works
The principle of operation is illustrated in
Fig. 1 . The electron beam produced by a
filament, cathode and associated gun
structure travels along the axis of the
TWT, before being collected by one or
more electrodes (collectors). The helical
circuit spaced closely around the beam
axis has a structure that causes it to pro-
pagate an RF wave that is slow with re-
spect to the speed of light. The helix
propagation velocity depends on the
power rating of the TW F, and is typically
10-30% of the speed of light. An input
cavity is provided to couple the RF signal
to the 'slow' wave structure. The ampli-
fied RF output signal is similarly taken
from a cavity.
The collector voltage and filament
emission are accurately controlled so that
the velocity of the electron stream is ap-
proximately the same as the axial phase
velocity of the RF input wave on the cir-
cuit. If the helix is properly proportioned,
its phase velocity is almost independent of
frequency over a wide range. It is, there-
fore, not uncommon for a TWT to have a
bandwidth of more than an octave.
The electron stream is density-modu-
lated because the longitudinal component
of the field generated by the 'slow' wave
interacts with the electrons travelling in
approximate synchronism with it. The re-
sult of the modulation is that the electron
stream induces additional waves on the
helix. Thus, along the length of the tube, a
portion of the direct-current energy of the
Frequency
Range (GHz)
Output
power (W)
1
Mass (kg)
Type number
Manufacturer)
Radio band 1
2.5 to 8
500
4.5
500CW
Teledyne
8
3.5 to 12
30
0.68
QKW5004
Raytheon
S
3.7 to 4.2
10
0.68
TL4010
AEG
S
4.5 to 10
1.5
0.9
191 078
EEV
c
1 7.9 to 8.4
60
-
N10025
EEV
.—2
6 to 18
40
0.68
QKW5005
Raytheon
— ~
8 to 18
2
0.7
191 0024
EEV
J
12 to 12.8
20
0.7
TL12019
AEG
Ku
14 to 14.5
200
3.2
Ku200W
Teledyne
Ku
29 to 31
12
1
TL30011
AEG
Ka
Table 1. Across the spectrum spread: listing of ten TWTs capable of working at different
bands in the radio frequency spectrum.
1 2.60 elektor india december 1989
Fig. 2. Typical relative TWT power gain as
a function of accelerating voltage.
electron stream is transferred to the circuit
as RF energy, resulting in amplification of
the RF input wave.
The all-important synchronism be-
tween the electron beam and the RF re-
quires accurate control of the accelerating
voltage, which is by no means simple to
implement in a spacecraft. The graph in
Fig. 2 shows the typical dependency of the
RF power gain on the beam accelerating
voltage.
Magnetic focusing
In order to control the physical size of the
electron beam in a TWT a focusing field is
required, providing a strength that en-
ables the charge forces to be compensated
that would otherwise cause excessive
beam divergence. The need of weight and
size reductions in satellites have forced
the development of permanent-magnet
focussing structures in which the field is
reversing periodically. Owing to various
technical limitations, electrostatic focus-
ing has not (yet) proved a viable alterna-
tive to magnetic focusing.
A carbon-based attenuator structure is
often fitted along the beam axis to enhance
the stability of the TWT (at gains of more
than 50 dB, oscillation is a real hazard).
THEORETICAL BACKGROUND
TO TRAVELLING-WAVE TUBES
The electron velocity, v, in cm/s is a func-
tion of the accelerating voltage, V, as ex-
pressed in
i> = 5.93 x10 V*
The approximate power gain, G, in deci-
bels, of a TWT may be calculated from
G = A + BCN
where
A is the initial mode establishing loss on
the helix. Typical values are -6 dB to
-9 dB;
B is a gain coefficient representing circuit
attenuation and space charge:
C is a gain parameter determined by the
impedances of the circuit and the elec-
tron stream;
N - the number of active wavelengths in
the tube.
Factor C is accounted for by
C
and N by
'a
, 2 P x ( )
(co/v) 2 8 Vo
N = ( l/X o ) (c/v )
where
to = beam current
Vo = beam voltage
/ = axial length of the helix
Xo = free-space wavelength
v = phase velocity of wave along tube
c = speed of light.
Voltages and currents
To obtain maximum efficiency from a
TWT, its operating voltages are all-im-
FREQUENCY «CHll
Fig. 3. Typical TWT small-signal gain
characteristics.
portant. There are 3 main voltages to con-
sider: the collector voltages, the helix volt-
age, and the heater voltage. Table 2 list the
voltage and current specifications of a
number of TWTs.
Collector voltages are usually of the
order of 2 kV, although the current trend
is towards voltages below' 1 kV. Collector
current is typically between 20 mA and
1 A. Voltage regulation to within 10% is
required for reasons outlined above.
Multiple collectors can help to increase
efficiency.
Helix voltages are typically between
2 kV and 10 kV, and currents between
10 mA and 500 mA.
The heater voltage, finally, is between
3.5 V and 6.3 V at a current demand of
0.5 A to 2.5 A. The filament heats up the
cathode to a temperature of about 650 °C
to enable electron emission to take place.
Type
»
Voltage (kV)
Current (mA)
Collector 1
Collector 2
Helix
Heater (V)
Cathode
Collector
Helix
Heater (A)
Efficiency
(%>
Gain (dB)
500CW
4.2 .
2.2
6.3
650
65
3.4
QKW5004
1.45
2.5
6.3
135
55
TL4010
1.55
37
40
N1078
2
2
25
37
N10025
2.1
49
34
28
QKW5005
1.8
3.8
6.3
135
12
0.5
40
N1024
2.5
2.5
22
TL12019
4.2
44
37
Ku200W
8.6
6.3
215
3
1.4
TL30011
5
38
29
Table 2. Electrical characteristics of a selection of TWTs.
elektor India decomber 1989 1 2.61
Special applications and
developments
Pulsed TWTs have been developed to pro-
duce a short coherent burst of RF energy,
for radar applications. The frequency,
bandwidth and peak-power specifica-
tions of these special TWTs have been op-
timized to meet the demands of radar
users.
Modem metallurgical processes have
enabled TWTs to be produced with a low
mass and special alloy focusing magnets
that give accurate beam control. Low' mass
of the TWT and, of course, its associated
multi-voltage power supply, are prime
considerations to keep the payload w'eight
of launch vehicles to a minimum.
What to look forward to
Recent history has seen industry commit-
ment for delivery of amplifiers that cover
the frequency range of 10,7 GHz to
12.7 GHz, mainly as a result of the increas-
ing use of satellite-TV in the communica-
tions and direct-broadcasting segments of
the X and Ku radio bands. Tube designs
that can address this whole bandwidth are
in the inventory of a number of major
TWT manufacturers including Telefun-
ken, Varian Associates, T-CSF and
Hughes EDD. It is important, however, to
recognize that new circuit technologies
M W PRODUCTS
Electrostatic Film Cleanser
Circuit Aids Inc introduces Electrostatic
Film Cleanser indigenously manufac-
tured meeting to International Stan-
dards.
This instrument, solves the film cleaning
problem eliminates static charges and
dust and other impurities permanently.
It features single pass operation with no
contamination with total static control.
Widely used in photographic films,
laminators, PBC manufacturers, etc.
based on 2-stage collectors are showing
promise of efficiencies previously associ-
ated only w'ith 4-stage collector designs.
In addition, these 2-stage collector de-
signs are expected to yield substantially
improved phase linearity over 'classic' de-
signs and could, to a large extent, help to
remove, or at least relax the requirements
of, linearization devices from future TWT
systems.
Research has shown that a typical Ku-
band satellite-TV TWT with a bandwidth
of 2 GHz and a 2-stage collector may be
expected to exhibit greater than 50% effi-
ciency w'ith a 4-stage depressed collector.
The previously mentioned developments
in TWT technology, however, allow de-
vices to be produced that provide efficien-
cies up to 54% with 2-stage collectors. In
these new TWTs, the 2-stage collector has
not been modified. The circuit improve-
ment, which primarily involves optimiza-
tion of velocity taper techniques,
produces beam efficiencies of the order of
27-30%, which is significant at X and Ku-
band frequencies. In addition, these new
circuits further reduce phase distortion
with typical AM-PM conversion at 2 to
4 dB. Also, third-order intermodulation
(IM) products are significantly reduced.
At saturation, the two-carrier third-order
IM product is not less than 14 dB down
from single-carrier saturation.
In conclusion, it is interesting to project
M/s. Circuit Aids Inc. • No. 451, II floor,
64th Cross • V Block • Rajajinagar •
Bangalore- 560 010. Tel: 359694.
Voltage Spike and Noise
Suppression Outlet Strip
Magnum have developed a voltage spike
and noise suppression outlet strip called
SPIKEBUSTER for computers, compu-
ter peripherals, audio equipment, TVs,
CTVs, VCRs, VCPs, copiers, medical
equipments, laboratory instrumenta-
tion, communications systems, photo-
composing machines, programmable
logic controllers and other devices con-
taining sensitive integrated circuits and
electronic tubes.
Consisting of an EMI/RFI filter and a
voltage spike protector circuit built into
a power strip with three 5 amp sockets/
one 15 amp socket. An OEM version
providing the output on a 15A 3-crore
cable in lieu of the sockets is also availa-
ble.
It prevents sensitive electronic equip-
ment from malfunctioning severely or
being badly damaged on account of
specific disturbances on the electricity
mains.
nnouticr
Fig. 4. Typical TWT saturated power out-
put as a function of RF input frequency.
the' performance, and in particular the ef-
ficiency, of TWTs that utilize these new
techniques w'ith 3 or 4-stage collectors.
Conservative estimates would place mini-
mum TWT efficiency at 58 to 60% for the
next generation of low-mass devices.
Magnum Electric Company Pvt. Limited
• 2, Ramavaram Road • Manapakkam •
Madras- 600 089.
Know-How for the Manufacture
of Electronic Chokes
Craftsman Electric is offering know-how
for the manaufacture of Energy Saving
Electronic Chokes used in Tube-Lights
of 40 Watts capacity (4 feet).
Electronic Chokes have the advantage.
Low Power Consumption,
Longer Tube Life,
Low voltage operation.
Produces less heat generation
No starter bulb or capacitor required.
Improved Power Factor, and
Better illumination.
Craftsman Electric • 149, West Samban-
dam Road • Coimbator-641 002. Tamil
Nadu.
12.62 efektor indis december 1989
NEW PRODUCTS
COMPONENT SOCKET ADAPTORS
& COVERS
These are suitable for mounting discrete
components such as resistors,
capacitors, diodes and other electronic
components, forming into a circuit of re-
quired design, and are designed to plug
directly into IC sockets as modular parts.
These carriers conserve space on PC
board by enabling maximum density of
packaging. Contact rows are spaced at
0.300” & 0.600” centres. The contacts
are spaced at 0.100” & 0.200” centers.
These are available in various sizes from
2 to 40 pins. Top covers which can be eas-
ily glued to the adapters are available for
8,14,16 & 24, 40 pins. These covers pro-
tect the circuit. These devices are used
for assembling modular & subminiature
circuits and also in microprocessors as
programmable shorting plugs.
Instrument Control Devices • B-4,
Abubaker Compound • Behind Garib
Nawaz Hotel • Raghvendra Mandir
Road • Oshiwara • Bombay -400 102.
TRUCK INDUCTIVE PROXIMITY
SWITCHES
HANS TRUCK GmbH & Co. KG,
West Germany, manufacture Inductive
Proximity Switches with sensing distance
of 60 mm, based on the principle that the
current in an oscillator circuit is altered
when metal enters or leaves its oscil-
latiang field. The oscillator coil is built
into a ferrite core and an H.F. magnetic
oscillating field is produced at the active
face of the switch. Metal entering the
field damps the oscillator and reduces
the current drawn by the oscillator cir-
cuit. The current change is used to pro-
vide switching signal. Oscillation nearly
ceases when the active face is fully co-
vered by metal.
These products can be imported under
OGL.
M/s. Arun Electronics Pvt. Ltd. • B/125-
126, Ansa Industrial Estate • Sakivihar
Road • Sakinaka • Bombay-400 072. •
Tel: 583354/587101.
FOUR PORT SERIAL CARD FOR
XENIX/UNIX
Mega’s MTS 8903 Four Post Serial Card
iis an interface to connect upto 4 termi-
nals to any IBM compatible PC/AT286/
AT386 running under Unix operating
systems. Compatible with the AST 4
post card , the MTS 8903 has four RS 232-
C asynch-ronous serial ports. The card II
O address and the interrupts are selecta-
ble. Further two of the ports can be con-
figured as standard PC serial ports.
The MTS 8903 Four Port Serial Card can
also be used under MS-DOS with the
support of a device drive to perform file
transfer and device sharing between 4
PC/XT/AT and a host computer which
may also be a PC/XT/ AT.
M/s. Mega Tromech Systems Pvt. Ltd. •
24, 12th Main • 1st Block • Rajajinagar •
Bangalore- 560 010.
DIGITAL IC TESTER
Features:
Function table of any Digital IC can be
checked within seconds without any ex-
ternal wire and soldering.
Total sixteen thumb-wheels are pro-
vided for easy programming.
An imported zip IC socket is provided
for easy fixing and removal of ICs.
A rectangular current meter to measure
the current drain by the IC.
Built-in regulated Power Supply.
An independent IC7447-cum-BCD tes-
ter.
Five logic level indicators are provided
for monitoring the out-put.
Leptron Electronics Products • 8, Vid-
hyanagar • JALNA-431 203.
elekJor india december 1989 12.63
NEW PRODUCTS
MODULAR PCB MOUNTING
MULTIWAY TERMINALS
These are specially designed for Elec-
tronic Printed circuit boards. These con-
nectors are available in 2 way & 3 way
lengths and can be interlocked into each
other to form required number of ways
with 5 mm pitch distance conforming to
international standards. The connection
is by soldering of pins on the printed cir-
cuit boards, and screw clamping the wire
termination. The housing is moulded in
special industrial grade plastics. The
maximum wire size is 2.5 mm square and
rated for 10A current.
Instrument Control Devices • B-4,
Abubaker Compound • Behind Garib
Nawaz Hotel • Raghvendra Mandir
Road • Oshiwara • Bombay-400 102.
SINGLE PHASE DC POWERPACKS
Static Power Systems offers Megacorp
Single Phase Powerpacks, manufactured
with technology using Thyristor Control ,
these power packs are primarily used for
speed control of DC Motors provided in
Plastic extruders, Printing, Rubber and
Type Machineries, Welding equip-
ments, Packaging machines etc. Single
phase powerpacks are available up to 5
H.P. (3.7 KW) Ratings and can also be
used as basic convertor in manufacturing
battery chargers, Electroplating re-
ctifiers, Power Control units for ovens.
Regulated DC power supplies etc.
Megacorp Power Packs are also availa-
ble in three phase versions up to 200 KW
Ratings and are made of Expoxy coated
chasis which can be readily mounted by
various OEM’s in the main panels of
their machines.
M/s. Static Power Systems Pvt. Ltd. • D-
148, Bonanza Indl. Estate • Ashok Chak-
ravarty Road • Kandivali (East) • Bom-
bay-400 101.
Bulk Requirements of ICs of Vari-
ous Types
Cycl-O Computers is an importer and
stockist of RAM, dynamic RAM, bipo-
lar PROMs; op-amps, voltage com-
parators, voltage regulators, line receiv-
ers, peripherals drivers, memory driv-
ers, display drivers; TTLs-LS, S, H,
ALS, AS, HC. PALs, remote servo con-
trollers, remote controls, transmitters
and receivers; photo detectors, LED dis-
plays, fiber optic components, source
and detector, assembly, opto couplers
isolators, 8-bit/16-bit microprocessors,
A/D and D/A converters, analogue
switches, amplifiers, counter circuits,
clock circuits, discretes/FETs, display
drivers, data communication, linear de-
vices, multiplexers, ROMs/EPROMs,
microprocessors and pheripherals; and
power transistors TO220, T03, fast
switching transistors, fast swiching dar-
lington, diodes, zeners, thyristors and
triacs/diacs.
For more details write to:
Cycl-O Computers • 308 Diamond Plaza
• Above Swastik Cinema * Lamington
Road • Bombay- 400 004.
SOLID STATE RELAYS
Satronix have introduced PCB/Chassis
Mount solid state relays of 2A, and
Amps output current relays. AC input
models can accept AC signals ranging
from 90 to 280 Volts AC. The DC input
signals can be operated from 3 to 32
Volts making them easy to interface to
the microprocessors and other logic level
devices. Output voltage can be selected
from 40 to 280 Volts AC. All relays in the
series benefit from zero-voltage turn on
as well as the zero current turn-off.
Solid-State design without mechanical
contacts and associated arcing virtually
eliminates electro magnetic interference
and transients.
Satronix • Module • 1 Electronic Sadan
1, • Bhosari MIDC • Pune-411 028.
12.64 elektor india december 1989
R. N. No. 39881/83
Allowed to post without prepayment LIC No. 91
MH BY WEST-228
LIC No. 91
PRECISION
WE MAKE
PERFORMANCE
OP-AMPS
AFFORDABLE
The AD 707 features the best
d.c. accuracy specification
available in a non-chopper
stabilized design, it features
a maximum Input offset
voltage of 1 5 pV (C Grade) &
input offset voltage drift of
0.1 pV/°C (C Grade)
The AD 548/648 features ultra
low input bias current-down
to 10 p A.
The AD 707/548/648 are
available in the plastic MiNl-
dip, cerdip & TO-99 metal
can. The AD 707 Is also
available in an 8 pin plastic
small outline (SO) package.
AD707JN AD548JN AD 648 JN
(Single)
(Dual)
input Was
current
2-SnA
20DA
20pA
input offset
voltage
90 pV
2mV
2mV
input offset
voltage drtft
1.V/°C
20 nV/°C 20t*V/°C
input voltage
Noise p-p
0.6 nV
2eV
2*V
Price
(100'S)
$ 1.37 $ 0.82 $ 1.37
SPEED
The AD 744 is fast settling
BiFET op-amp. it can settle
to 0.01% (for 10V step) in 500
nsec.(K grade) and to 0.0025%
(for 10V step) in 1.5 psec (K
grade). It also has a slew rate
of 75 v/psec.
The AD 711/712 combines
good speed and bias current
specifications.
The AD 744/711/712 are
available in the plastic MINI-
OIP, CERDIP, and TO-99 metal
can.
AD744JN AD 711 JN AD712JN
(Single)
(Dual)
input Dias
current
100 PA
50 pA
75 pA
input offset
voltage
imV
2 mV
5 mV
Setting Time
to 0 . 01 %
0.9*5
1*5
1*5
Typical slew
rate
75 V/*S
20V/*S
20 V/^S
Price
MOO'S)
$ 2.47 $ 0.88 $ 1.37
Whether It is
Precision or speed
you can count on
the leader -
Analog Devices inc
For more details contact your nearest
Analog Devices representative in India
(kO ^' 648
ANALOG SALES (INDIA) PVT. LTD.
Pune (REGO OFF) : 149.1 -A Plot No 5 Krishna. Aundh. Pune 411 007. Ph : 53880 TLX 145-470
N Delhi (BR. OFF) : C-197 Sarvodaya Enclave. New Delhi 1 10017. Ph 6862480 TLX 031-73228
Bangalore (BR. OFF) : 992 13th Main Rd. Indiranaflar. Bangalore 580038. Ph : 560506 TLX 845-8994
ANALOG
DEVICES
SJAS 8848